GitHub - XAIHT/Tlamatini: Tlamatini is a self-hosted AI developer assistant for local or cloud LLMs: hybrid RAG over your codebase, a 79-tool multi-turn engine, a 74-agent visual workflow designer, and ACPX, a runtime for external coding agents like Claude Code, Cursor and Gemini CLI.

She doesn't chat. She does.

"One who knows" — in Nahuatl.
A self-hosted AI developer assistant that turns your PC into a workshop where 76 agents collaborate to actually get things done.

▶️ Watch the one-minute demo  ·  🌐 Website  ·  🎬 More demos  ·  ⚡ Quickstart

🌟 What can you do with her? A few examples in plain English:

"Read this whole codebase, find every endpoint that touches the user table, and write me a security report."
"Open Chrome, log into our staging dashboard, screenshot the analytics page, email it to the team."
"Build the STM32 firmware, flash it to the connected board, and watch the serial output for errors."
"Spin up Claude Code, give it the refactor task, then hand the output to Cursor for testing."
"Design a flow that runs every morning, scrapes our competitors' pricing, and posts the diff to Slack."

She can do all of that — self-hosted on your own machine, powered by the model you choose: a local Ollama model, Ollama Cloud (the shipped default), or a cloud API like Anthropic Claude.

🚀 What it gives you that a plain chatbox doesn't

💡 The thesis behind Tlamatini

A 20B-parameter local LLM, given the right tools, the right agents, and skills fine-tuned to your operating procedures, outperforms a 200B cloud model with no tools — almost every time.

Parameter count isn't the bottleneck. The toolbelt is. Tlamatini is the toolbelt.

⚡ Quickstart in 5 minutes

Three steps: install Ollama, pull the models Tlamatini ships with, run Tlamatini from the release ZIP.

1 · Install Ollama (no admin rights — full detail in §2.2):

$env:OLLAMA_INSTALL_DIR = "$env:LOCALAPPDATA\Programs\Ollama"
irm https://ollama.com/install.ps1 | iex
ollama serve     # leave this running in its own window

2 · Pull the six default models — these exact tags, nothing else (the five :cloud tags need an Ollama Pro/Max plan, see §2.4; only Nomic-Embed-Text:latest runs locally):

ollama pull glm-5.1:cloud
ollama pull gpt-oss:120b-cloud
ollama pull qwen3.5:397b-cloud
ollama pull qwen3.5:cloud
ollama pull kimi-k2.6:cloud
ollama pull Nomic-Embed-Text:latest

3 · Install Tlamatini from the release ZIP — download Tlamatini v1.19.3, then:

1. Unzip the release archive anywhere (no admin rights needed).
2. Run  Installer.exe  → pick an install directory → finish.
   (The bundled Python 3.12.10, Java, Git and Playwright browsers are carried in —
    nothing else to install.)
3. Launch Tlamatini from the desktop / Start-Menu shortcut.
4. Open  http://127.0.0.1:8000  and log in with   user / changeme   — then say hi.

Prefer running from a cloned repo instead of the ZIP? Use the full §2 Quickstart (source mode).

📖 Long-form docs  ·  🏷️ Versioning  ·  🤝 Contributing

Table of Contents

• 1. Overview
  • 1.1. What Tlamatini is
  • 1.2. What it gives you that a plain chatbox does not
  • 1.3. Demo videos
• 2. Quickstart (source mode)
  • 2.1. Prerequisites
  • 2.2. Install Ollama (no admin rights)
  • 2.3. Pull the default models
  • 2.4. Cloud models require an Ollama Pro/Max plan
  • 2.5. Clone, install, migrate
  • 2.6. Run the server (not-frozen)
  • 2.7. Log in for the first time
• 3. Using the Chat (/agent/)
  • 3.1. Chat layout in 30 seconds
  • 3.2. Setting code as context
  • 3.3. Tutorial: a one-shot question (no toggles)
  • 3.4. Tutorial: the internet toggle
  • 3.5. Tutorial: the Multi-Turn toggle
  • 3.6. Tutorial: the Exec Report toggle
  • 3.7. Tutorial: the ACPX toggle
  • 3.8. Tutorial: the Ask Execs toggle
  • 3.9. From chat to flow: the Create Flow button
  • 3.10. Why Chat-created flows are safer now
  • 3.11. The DB menu — Backup, Set DB, and the start-up swap-in
  • 3.12. The ACPX-Skills menu — Browse, Configure, Diagnostics, Reload
  • 3.13. Tutorial: command a window from chat (chat_agent_windower)
  • 3.14. Tutorial: drive a browser from chat (chat_agent_playwrighter)
  • 3.15. Tutorial: run Kali Linux tools from chat (chat_agent_kalier)
  • 3.16. Tutorial: build and flash STM32 firmware from chat (chat_agent_stm32er)
  • 3.17. Tutorial: build and upload ESP32 firmware from chat (chat_agent_esp32er)
  • 3.18. Tutorial: build and upload Arduino firmware from chat (chat_agent_arduiner)
• 4. Visual Workflow Designer (/agentic_control_panel/)
  • 4.1. Canvas anatomy
  • 4.2. Tutorial: your first flow (3 agents)
  • 4.3. Saving and loading .flw files
  • 4.4. Validate and Start now compile the live canvas
  • 4.5. Pause, Resume, Stop
  • 4.6. FlowHypervisor (watchdog)
  • 4.7. FlowCreator (let an LLM design the flow)
  • 4.8. Parametrizer (chain outputs into the next agent's config)
  • 4.9. Gatewayer (external triggers)
• 5. ACPX — External Coding-Agent CLIs as Tools
  • 5.1. What ACPX is
  • 5.2. Supported agent_ids and transports
  • 5.3. The 12 ACPX/Skill tools
  • 5.4. Tutorial: spawn-and-go (single agent)
  • 5.5. Tutorial: multi-CLI relay with acp_relay
  • 5.6. API key setup (the easy button)
  • 5.7. ACPXer — the visual canvas counterpart
• 6. Unreal MCP — Driving Unreal Engine 5 from Tlamatini
  • 6.1. What Unreal MCP is
  • 6.2. The MCP plugin source (the MCP git location)
  • 6.3. Installing and enabling the plugin inside your UE5 project
  • 6.4. The command catalog (up to 53 commands across 9 categories)
  • 6.5. Using Unreal MCP from the chat (chat_agent_unrealer)
  • 6.6. Using Unreal MCP on the canvas (the visual Unrealer node)
  • 6.7. What the agent actually does, end-to-end
  • 6.8. Exec Report integration
  • 6.9. Bullet-proof checklist for Unreal Engine users
  • 6.10. Troubleshooting Unreal MCP
• 7. Building a Frozen Distribution
  • 7.1. Three-step pipeline
  • 7.2. Step 1 — build.py
  • 7.3. Step 2 — build_uninstaller.py
  • 7.4. Step 3 — build_installer.py
  • 7.5. What the installer does on the end-user box
  • 7.6. Source mode vs Frozen mode: why flows still work
• 8. Configuration (Tlamatini/agent/config.json)
  • 8.1. LLM and unified-agent
  • 8.2. RAG
  • 8.3. ACPX
  • 8.4. MCP services and other knobs
• 9. Architecture at a Glance
  • 9.1. Big picture
  • 9.2. The five layers
  • 9.3. Multi-Turn execution pipeline
  • 9.4. Agent contracts and the Flow Compiler
  • 9.5. Agent catalog (the 76 types, by family)
  • 9.6. Self-Knowledge & Self-Modification
• 10. Embedding-Memory Pre-Flight Guard (GPU hosts)
  • 10.1. Why this exists
  • 10.2. How the check fires (the hook point)
  • 10.3. GPU detection and no-GPU behavior
  • 10.4. Three-tier VRAM prediction
  • 10.5. The 80% threshold and the chat-bubble warning
  • 10.6. Tuning, overrides, and what the guard does NOT do
  • 10.7. Test coverage
• 11. Orphan-Process Cleanup (conhost.exe reaper)
  • 11.1. The problem this solves
  • 11.2. The three tiers
  • 11.3. What gets reaped (and what does not)
  • 11.4. The user-visible follow-up message
• 12. Troubleshooting
  • 12.1. Ollama / models
  • 12.2. RAG / context
  • 12.3. Multi-Turn / planner
  • 12.4. Chat-created flows and ACP validation
  • 12.5. ACPX / external CLIs
  • 12.6. Frozen build / installer
  • 12.7. Logs to consult first
• 13. Versioning
  • 13.1. The bump rules
  • 13.2. Cutting a release
  • 13.3. Where you can see the running version
  • 13.4. Building without tagging (development)
  • 13.5. Overriding the resolved version
• 14. Contributing & License
  • 14.1. Contributing
  • 14.2. Acknowledgments
  • 14.3. License

1. Overview

1.1. What Tlamatini is

Tlamatini (Nahuatl for "one who knows") is a Django/Channels app you run on your own machine. It packages a hybrid RAG pipeline, a Multi-Turn tool-calling LLM loop, an ACPX runtime that spawns external coding-agent CLIs as child processes, an Unreal MCP client that drives Unreal Engine 5 from chat or canvas, and a drag-and-drop workflow designer with 76 agent types — into one self-hosted install. The app and RAG run locally; the LLM backend is configurable: Ollama (local models or Ollama Cloud — the shipped default uses Ollama Cloud :cloud tags), Anthropic Claude (cloud), Qwen vision (Ollama).

License: GPL-3.0 · Repo: https://github.com/XAIHT/Tlamatini.git · Platform tested: Windows 11 (cross-platform for source mode).

1.2. What it gives you that a plain chatbox does not

1. Real RAG over your code — FAISS + BM25 hybrid retrieval, code-aware metadata extraction, Reciprocal Rank Fusion, context budgeting, OOM fallback.
2. Multi-Turn mode — the LLM becomes an operator: shell, Python, APIs, SQL, file ops, screenshots, keyboard/mouse automation, email, Telegram, WhatsApp, STM32 firmware build/flash — chained in one conversation.
3. ACPX — delegate sub-tasks to external CLIs (claude, cursor-agent, codex, gemini, qwen-code, plus 8 more) and relay output between them.
4. Visual workflow designer — design .flw flows once, run them unattended, schedule with Croner, watch them with FlowHypervisor.
5. Self-aware — a first-person self-knowledge map (Tlamatini.md) is injected into the LLM's prompt on every chain, so Tlamatini can answer accurately about her own architecture, runtime modes, ports, and pages. Builds packaged with --self-modify ship her own source tree (TlamatiniSourceCode/) so she can read, inspect, and modify herself.

Everything runs locally. The whole app packages into a one-click Windows .exe distribution (Part §7).

1.3. Demo videos

• First system-usage walkthrough
• Loading a complete project and summarizing its source code
• Installing OpenCV end-to-end in Multi-Turn
• Uninstalling Poco — Exec Report and matching flow
• Implementing a FlowCreator-aided agentic flow
• A complete Cybersec enhancement with Tlamatini!!!

2. Quickstart (source mode)

This is the fastest way to be productive: clone, install, run. No installer, no admin, no frozen build. Five minutes.

2.1. Prerequisites

You do not need administrator rights for any of the steps below.

2.2. Install Ollama (no admin rights)

Open PowerShell normally (do not Run as Administrator), then:

$env:OLLAMA_INSTALL_DIR = "$env:LOCALAPPDATA\Programs\Ollama"
irm https://ollama.com/install.ps1 | iex

Close the window, open a fresh PowerShell, and verify:

ollama --version
ollama serve     # leave running in its own window if it's not already up
Invoke-WebRequest http://127.0.0.1:11434/api/tags -UseBasicParsing

Tlamatini expects Ollama at http://127.0.0.1:11434.

2.3. Pull the default models

ollama pull Nomic-Embed-Text:latest
ollama pull kimi-k2.6:cloud
ollama pull qwen3.5:cloud
ollama pull gpt-oss:120b-cloud
ollama pull qwen3.5:397b-cloud
ollama pull glm-5.1:cloud

You can substitute any tag — just edit Tlamatini/agent/config.json (see §8.1) or the relevant agent's config.yaml.

Optional: swap to a higher-detail embedding model. If your retrieval quality on dense, technical corpora is not good enough with the default, you can switch to qwen3-embedding:8b from the Config → Models menu inside the app (or by editing embeding-model in config.json and reconnecting). Use with caution: qwen3-embedding:8b is roughly 10× heavier in VRAM than Nomic-Embed-Text:latest (~6.24 GB resident vs ~600 MB on a Q4_K_M quant) and will trip the embedding-memory pre-flight guard (see §10) on 8 GB consumer GPUs. Pull it first with ollama pull qwen3-embedding:8b.

2.4. Cloud models require an Ollama Pro/Max plan

Five of the six default model tags in §2.3 carry the :cloud suffix — kimi-k2.6:cloud, qwen3.5:cloud, gpt-oss:120b-cloud, qwen3.5:397b-cloud, and glm-5.1:cloud (only Nomic-Embed-Text:latest runs locally). Those are Ollama Cloud models: they live on Ollama's servers, not on your machine, and ollama pull only registers a stub that proxies inference to the cloud. Reaching that cloud requires a logged-in Ollama account and a subscription tier that allows the workload you intend to run.

The plan structure (prices are deliberately omitted from this README because they change — check https://ollama.com/pricing for the current numbers):

If you do not want to subscribe, you can run Tlamatini entirely on local open-weights models. Edit Tlamatini/agent/config.json (chained-model, unified_agent_model, mcp_file_search_model, flow_creator_model, image_interpreter_model) and every agent config.yaml that names a :cloud tag, and swap them for a model you have pulled locally (for example, llama3.1:8b, qwen2.5-coder:14b, mistral-nemo:12b). Performance and quality will scale with your GPU/CPU — Multi-Turn and ACPX both work fine on a sufficiently large local model.

API keys are separate. This subscription only governs *:cloud Ollama models. The ACPX runtime can additionally spawn external coding-agent CLIs that bring their own credentials (Anthropic API key for claude, OpenAI key for codex, Google key for gemini, etc.) — those are configured in Tlamatini/agent/config.json under acpx.agents.<id>.env and are unaffected by your Ollama plan. See §5.6 for the easy-button setup. (Unreal MCP is not part of ACPX — it's its own MCP surface, documented in §6.)

2.5. Clone, install, migrate

git clone https://github.com/XAIHT/Tlamatini.git
cd Tlamatini

python -m venv venv
# Windows:
venv\Scripts\activate
# Linux/macOS:
source venv/bin/activate

pip install -r requirements.txt

python Tlamatini/manage.py migrate
python Tlamatini/manage.py createsuperuser
python Tlamatini/manage.py collectstatic --noinput

2.6. Run the server (not-frozen)

python Tlamatini/manage.py runserver --noreload

--noreload is important: Daphne's auto-reloader does not coexist well with the wrapped-runtime subprocess pool.

The console title becomes Tlamatini and stdout/stderr are tee'd into Tlamatini/tlamatini.log (truncated on every start). When debugging, tlamatini.log is the first thing to read.

2.7. Log in for the first time

Open http://127.0.0.1:8000/ and log in with the superuser you just created. Then:

• /agent/ — the chat (Part §3)
• /agentic_control_panel/ — the visual designer (Part §4)
• /admin/ — Django admin (change passwords, manage users)

If you used the installer (Part §7) instead of cloning, the default credentials are user / changeme. Change them at first login via /admin/.

3. Using the Chat (/agent/)

3.1. Chat layout in 30 seconds

┌───────────────────────────────────────────────────────────────────────────────┐
│ Tlamatini  [Context ▼] [Open in… ▼] [MCPs ▼] [Tools ▼] [Agents ▼] [Config ▼] [DB ▼] │ ← top nav
├───────────────────────────────────────────────────────────────────────────────┤
│  Multi-Turn ☐  Exec Report ☐  ACPX ☐  Ask Execs ☐  internet ☐   Clear ⌫     │ ← five toggles
├───────────────────────────────────────────────────────────────────────────────┤
│  ┌──── chat ────────────────┐   ┌──── code canvas ────────────────┐          │
│  │  conversation history    │   │  syntax-highlighted, with copy  │          │
│  └──────────────────────────┘   └─────────────────────────────────┘          │
├───────────────────────────────────────────────────────────────────────────────┤
│  Type your prompt here…                                              [Send]   │
└───────────────────────────────────────────────────────────────────────────────┘

The five toolbar toggles are independent (except Ask Execs, which only activates while Multi-Turn is on). Tick whatever combination fits the task — each one is its own tutorial section below.

Newer builds also expose a Config dropdown in the same navbar. Config -> Models edits the most common model-name fields, and Config -> URLs edits the Ollama / unified-agent / MCP endpoint values through validated dialogs instead of hand-editing JSON. The chat/canvas divider was also tightened so resizing the right-hand canvas feels more predictable during long editing sessions.

The newest entry in that navbar is the DB dropdown: DB -> Backup database snapshots the live SQLite file to a directory you pick, and DB -> Set DB stages a db.sqlite3 file of your choice for the next session — Tlamatini swaps it in before Django opens the database, archives the previous one under DB/Older/<timestamp>/, then continues normal start-up. Full walkthrough in §3.11.

3.2. Setting code as context

Click Context in the top nav:

Set directory as context now loads a project at any depth under the app root. The old browser showDirectoryPicker() only exposed the leaf folder name, so deeply-nested projects could not be reached; it was replaced by a backend native Win32 folder picker (views.pick_context_directory_view, route pick_context_directory/) that returns the real absolute path. path_guard.is_within_application_root() then accepts the application root or any descendant of it, and agent_page_init.js falls back to manual path entry on non-Windows hosts.

A green banner at the top shows the current context path. If embedding runs out of memory, Tlamatini packs the source files as a fallback context — retrieval quality drops, access to your code does not.

If you refresh the browser and Tlamatini restores a saved context automatically, the input now stays disabled until the contextual RAG chain has actually finished rebuilding. That closes the old "restored banner arrived before the context was really ready" race on the first load stage.

3.3. Tutorial: a one-shot question (no toggles)

Leave every checkbox unticked. Type:

"Write a Python function that validates an email address with a regex. Just the function."

The bot answers in one shot. Code lands in the right-hand canvas with copy/save buttons. This is the legacy chat path — fast, no tools, no internet.

3.4. Tutorial: the internet toggle

Tick internet when the question genuinely needs fresh web data:

"What is the latest stable version of FastAPI right now?"

Tlamatini classifies the prompt with a small LLM call ("does this need the web?"), then DuckDuckGo-searches, summarizes the top results, and inlines the summary into the LLM's context. Leave it unticked for everything else (the round-trip adds latency).

3.5. Tutorial: the Multi-Turn toggle

This is the big one. Multi-Turn turns Tlamatini from answerer into operator:

• The planner picks the relevant subset of Tlamatini's 79 Multi-Turn tools — 20 core Python tools (execute_command, agent_starter, googler, the image-analysis pair, the chat_agent_run_* lifecycle helpers, …), 47 wrapped chat-agent tools, and 12 ACPX/Skill tools — binding at most max_selected_tools per request (default cap: 20).
• The unified-agent loop runs up to 4096 iterations (the unified_agent_max_iterations default) — call tool, see result, decide next, chain.
• Wrapped sub-agents run in headless background runtimes (no console pop-ups).

Try this: tick Multi-Turn, send

"Take a screenshot of my desktop and save it to C:\Tlamatini-test\shot.png."

Watch the chat. The LLM picks chat_agent_shoter, calls it with the right args, reads the JSON result, and replies "Done — saved to C:\Tlamatini-test\shot.png." Open the file. The screenshot is there.

Multi-Turn stacks with Set-Context: the LLM reasons over your code and runs tools on the result.

3.6. Tutorial: the Exec Report toggle

Below the prose answer, Tlamatini appends per-agent execution tables — one HTML table per kind of state-changing agent that fired. Each row = one real tool call + ✓/✗.

Tick Multi-Turn + Exec Report and send:

"Create C:\test\hello.txt with Hi from Tlamatini, then read it back and tell me its size."

After the prose, you see:

─── List of File Creator Operations ───
 #  │ Command                                        │ ✓/✗
 1  │ filepath='C:\test\hello.txt' content='Hi …'    │  ✓
─── List of Executer Operations ───
 #  │ Command                                        │ ✓/✗
 1  │ type C:\test\hello.txt                         │  ✓

What gets a table: state-changing tools only (execute_command, execute_file, unzip_file, decompile_java, every chat_agent_* that touches the system, all five acp_* lifecycle tools — merged into one "List of ACPx Operations" — and invoke_skill). Read-only tools (Crawler, Googler, Prompter, Summarizer, File-Interpreter/Extractor, Image-Interpreter, Shoter, Sleeper, monitor_, run_, window_present) are intentionally absent. Tables persist into chat history — reload the page and they are still there.

3.7. Tutorial: the ACPX toggle

ACPX lets the chat delegate to external coding-agent CLIs running on your box. Picture it:

You ─► Tlamatini chat ─► acp_doctor → acp_spawn(claude) → acp_send_and_wait
                                  │
                                  ▼ subprocess.Popen
                                claude CLI / gemini / cursor / codex / qwen / …

When ACPX is ticked, the planner sees the 12 ACPX/Skill tools. When unticked, those tools are filtered out — the chat behaves like legacy Multi-Turn. (Implemented in agent/acpx/__init__.py::filter_acpx_tools().)

Prereq: at least one external CLI on PATH. The simplest:

npm install -g @anthropic-ai/claude-code
claude --version

Then drop your key in Tlamatini/agent/config.json (or use the setup-new-acpx-key skill — much easier).

Tick Multi-Turn + ACPX + Exec Report and send:

"Use ACPX to spawn the claude CLI in C:/Development/Tlamatini, ask it to summarize CLAUDE.md in 5 bullet points, harvest the answer, and kill the session."

You see: acp_doctor (always first) → acp_spawn(agent_id="claude", task=…) → acp_send_and_wait → acp_kill. The 5 bullets land in the prose, and the Exec Report shows a "List of ACPx Operations" table with all four rows.

ACPX deep dive in Part §5.

3.8. Tutorial: the Ask Execs toggle

Ask Execs puts a human in the loop before every action. It only activates while Multi-Turn is on — the checkbox stays disabled and greyed until you tick Multi-Turn — because the prompt lives inside the Multi-Turn tool loop. When unticked, Tlamatini behaves exactly like the legacy Multi-Turn flow.

With Ask Execs ticked, before each state-changing Tool / MCP / Agent runs, the chain pauses and a modal dialog appears showing exactly what is about to happen:

• the Tlamatini Tool / MCP / Agent that is going to execute (e.g. Tool: Executer, Agent: SSHer, Skill: …),
• the underlying tool name,
• the parameters of execution (read-only textarea),
• the program to be executed (read-only textarea — the command / script / intent),
• the shell to be executed (read-only textarea — cmd.exe / PowerShell, Python interpreter, Remote SSH shell @ host, …).

Two choices:

• Proceed → that tool runs and the chain continues as normal to the next step (which prompts again).
• Deny → the entire chain stops immediately. No further tools run. You get back:
  • the prose answer so far,
  • the Exec Report tables of what did execute (only if Exec Report is also ticked),
  • and — always — a big red ⛔ "Execution interrupted" banner naming the exact Tool/MCP/Agent you denied, plus its program/command, shell, and parameters.

Read-only / polling tools (chat_agent_run_status, chat_agent_run_log, get_current_time, window_present, …) are not prompted — they only observe, they do not "execute".

Try this: tick Multi-Turn + Ask Execs (and optionally Exec Report) and send:

"Delete every *.tmp file under C:/Temp and then list what's left."

You'll get a permission dialog for the deletion step. Click Deny and the run halts with the red banner showing exactly the command you stopped; click Proceed and it carries on. This is the safety belt for destructive or sensitive operations — review each action before it touches your machine.

Implementation: the synchronous tool loop blocks on a browser round-trip via agent/exec_permission.py (ExecPermissionBroker). A denial fails safe; a missing browser never silently runs an unconfirmed tool. See §9.3.

3.9. From chat to flow: the Create Flow button

When a Multi-Turn run succeeds and used at least one state-changing tool, Tlamatini renders a Create Flow button on the message header. Click → download a .flw JSON file mirroring the exact tool sequence, laid out left-to-right, ready to load in the visual designer:

Starter ─► Crawler ─► File Creator ─► Ender

You can re-open it in /agentic_control_panel/ and run it as an unattended workflow. The LLM is no longer in the loop.

The button gates on four conditions: Multi-Turn was on, ≥1 mappable tool succeeded, an LLM-based classifier marked the answer SUCCESS (fails open on internal error), and the user is logged in.

3.10. Why Chat-created flows are safer now

Older Chat-created .flw files were generated almost entirely in the browser. That worked for simple chains, but it meant the browser had to remember many backend facts:

• what each agent is called on disk;
• which config field means "my input";
• which config field means "my output";
• which agents are special, like Ender or Parametrizer;
• which values are safe to save into a portable .flw file.

That is a lot of responsibility for a button.

Now the browser still builds the first draft, but the backend normalizes it through the Agent Contract Registry before the file is downloaded. In plain English: Tlamatini checks the flow against the same agent rules that ACP uses to run it.

What this means for you:

1. Repeated tools stay repeated. If Multi-Turn ran Executer five times, the flow contains five Executer nodes, not one overwritten node.
2. Agent names are normalized. Names like Kyber-KeyGen, kyber_keygen, and Kyber Keygen are resolved to the right template.
3. Secrets are protected where the contract knows about them. Remote chat super-agents such as TeleTlamatini and WhatsTlamatini have credential-like fields redacted on export.
4. The .flw file stays portable. It does not store C:/Development/... or the installed app path beside Tlamatini.exe.

If backend normalization is temporarily unavailable, the old browser generator remains as a fallback so the button does not become useless.

3.11. The DB menu — Backup, Set DB, and the start-up swap-in

The whole of Tlamatini — chat history, agents, Tool/MCP toggles, sessions, your user — lives in a single SQLite file. The DB dropdown gives you a safe, GUI-first way to handle that file: a read-only Backup path, a destructive-but-deferred Set DB path, and a built-in audit trail under DB/Older/.

DB -> Backup database

Opens a dialog with one input — the target directory. The path is live-validated (350 ms debounce): the page asks GET /agent/check_backup_directory/?path=… as you type and colors the status line green / amber / red:

Click Backup → Tlamatini calls POST /agent/backup_db/, resolves the live database path via settings.DATABASES['default']['NAME'] (so source / frozen behave identically), and shutil.copy2s it to <your-dir>/db.sqlite3. The live database stays open and unchanged.

DB -> Set DB

The opposite direction: replace the database on the next start-up. Same dialog idiom, stricter validation. The input is the full path to a db.sqlite3 file; the page asks GET /agent/check_set_db_file/?path=… as you type:

Click Set → POST /agent/set_db/ copies your file into <base>/DB/ToLoad/db.sqlite3 (where <base> is the executable directory in frozen mode, the inner Django project directory in source mode). The live database is not touched — SQLite is held open by Django, so the actual replacement must wait for a process restart.

Immediately after staging succeeds, the dialog is replaced by a yellow ⚠ warning panel with a single OK button:

The selected database will be loaded the next time Tlamatini starts. If you want it loaded immediately, you must restart Tlamatini completely so the swap-in can run BEFORE Django opens the live database.

If you click Cancel instead of Set, the staging dialog closes and nothing is written.

The start-up swap-in (the third, invisible leg)

The actual replacement lives at the very top of Tlamatini/manage.py, in _apply_pending_db_swap(). It runs before any Django import so Django's SQLite connection pool is never holding a stale file descriptor at the moment of the swap:

manage.py main()
    │
    ▼
_apply_pending_db_swap()
    │
    ▼
[ DB/ToLoad/db.sqlite3 exists? ]
    │
    ├─ NO  ──► return (no-op, normal start-up continues)
    │
    └─ YES ──► [1] mkdir DB/Older/<YYYY-MM-DD_HHMMSS>/
               [2] shutil.move(live db.sqlite3 → Older/<timestamp>/db.sqlite3)
               [3] shutil.move(DB/ToLoad/db.sqlite3 → live db.sqlite3)
               [4] return
    │
    ▼
from django.core.management import execute_from_command_line   ← only NOW Django wakes up

Three guarantees:

1. A Reconnect from the navbar is NOT enough. The swap window is only open before the Django process opens its SQLite pool. You must fully restart Tlamatini (close the console / kill the exe, then launch again).
2. Atomic moves, no copies. Both legs use shutil.move (filesystem rename when possible, copy+delete across mounts). A second launch with DB/ToLoad/ empty is automatically a no-op — no "stuck flag" to clear.
3. Mode-correct path resolution. Frozen mode reads <exe_dir>/DB/ToLoad/db.sqlite3 (where you can browse to it in Explorer); source mode reads <repo>/Tlamatini/DB/ToLoad/db.sqlite3 (next to manage.py). The live db.sqlite3 path is computed the same way Django does — _MEIPASS/db.sqlite3 under PyInstaller, <manage.py dir>/db.sqlite3 in source — so the swap-in always writes to exactly the path Django will open.

If anything fails inside the swap-in (locked file on Windows, corrupt source, permission error), the function catches the exception, prints --- [DB SWAP] Skipped due to error: … to tlamatini.log, and lets Tlamatini start normally with the previous database. A bad ToLoad file must never lock you out of your own database.

The Older audit trail

Every successful swap-in leaves a complete record under <base>/DB/Older/<YYYY-MM-DD_HHMMSS>/db.sqlite3. Because Set DB moves (not copies) the prior live database, this archive is the only built-in recovery path:

DB/
├─ ToLoad/                 ← empty most of the time; momentary home of next-session pick
│   └─ README.md
└─ Older/
    ├─ 2026-05-14_153022/db.sqlite3   ← was live before swap #1
    ├─ 2026-05-14_164410/db.sqlite3   ← was live before swap #2
    └─ README.md

To roll back, drop the archived db.sqlite3 back into ToLoad/ and restart — the swap-in will archive the current live database under a fresh timestamp and promote your roll-back pick. Tlamatini never auto-deletes anything from Older/; prune by hand when the tree gets noisy, but remember each file is a full snapshot of chat history + agents + sessions + your user.

Where the tree comes from

Both directories must exist on day one (the swap-in opens them with os.makedirs(exist_ok=True), but having them pre-seeded with docs prevents user confusion):

• Source / dev mode: Tlamatini/Tlamatini/DB/{ToLoad,Older}/README.md are checked into the repo. The README files are the "git keepers" — without them, git would silently drop the empty directories.
• Frozen mode: build.py extends its empty_dirs tuple with "DB/ToLoad" and "DB/Older". The PyInstaller post-build step creates both under dist/manage/, the pkg.zip packager preserves them via explicit zip entries, and end-users get the tree from the very first launch.

3.12. The ACPX-Skills menu — Browse, Configure, Diagnostics, Reload

Tlamatini ships with 27 skills — markdown SKILL.md packages under agent/skills_pkg/ that the LLM can invoke through invoke_skill('<name>', '{...args...}'). They cover everything from the canonical acp-router (pick the right external CLI for an intent) and summarize (compress text faithfully) to setup-new-acpx-key, skill-creator, flow-making (turn a plain-language objective into a canvas-loadable .flw by driving the FlowCreator engine — supersedes the legacy tlamatini-flow-from-objective), code-review (senior-engineer git-diff review with an APPROVE/REQUEST_CHANGES verdict), security-audit (multi-scanner SAST/secret/dependency sweep) and kali-pentest (an authorized Kali Linux assessment runbook that drives the Kalier agent / MCP-Kali-Server), the tlamatini_* audit/lint/refactor helpers, and integration stubs for GitHub / Notion / Slack / Gmail / Jira / Todoist / Trello / Weather.

Before 2026-05-17 the only way to interact with them was through the LLM (list_skills to enumerate, invoke_skill to run). The ACPX-Skills navbar dropdown — added next to Agents and Config in the chat toolbar — gives you an operator-grade admin surface that does NOT require the LLM. Four entries:

ACPX-Skills -> Browse Skills

Opens a two-pane modal: a left-side list of all 27 skills (with a green/red dot for enabled / disabled and a runtime tag) and a right-side detail pane that shows the selected skill's full identity — description, runtime (in-process / acpx), acpx_agent if any, budgets (max_iterations · max_seconds · max_tokens), trigger keywords, requires_tools and requires_mcps, inputs and outputs (with required-field markers), and the full markdown body. A search box at the top filters by name or description as you type. Pure read — nothing is written back.

Backed by GET /agent/skills/ (list payload) and GET /agent/skills/<name>/ (deep detail). Use it when you want to know what a skill actually does before you ask the LLM to call it, or when you've just authored a new SKILL.md and want to confirm it parsed correctly.

ACPX-Skills -> Configure Skills

A checkbox grid — one row per skill — that mirrors the existing MCPs and Agents dialogs exactly. Toggle a checkbox off, click Continue, and the row's Skill.enabled flips to false. Two consequences immediately:

• list_skills (the LLM's enumeration tool) filters that skill out of its returned array.
• invoke_skill('<name>', ...) returns {"ok": false, "code": "SKILL_DISABLED"} instead of running.

Toggling back to enabled restores the skill. This is the right knob when (a) you want to hide an unfinished skill from the planner, (b) you don't have the API key for an integration skill (e.g. notion without NOTION_TOKEN) and don't want the LLM to keep trying, or (c) you're running a demo and want a minimal tool surface.

The toggle goes over the same WebSocket channel as set-mcps / set-tools / set-agents — payload encoding name=description=true/false,name=description=true/false,.... Backend handler is in consumers.py::receive and calls save_skill(name, enabled) which touches only the enabled column.

ACPX-Skills -> Diagnostics

A cross-check report that catches drift between the skill catalog and the rest of the system. Sections:

• Missing tool dependencies — for each skill whose requires_tools lists a tool that's currently disabled in the Tools dialog, lists the skill + the unmet tools. (A disabled tool means the skill would fail at runtime — Diagnostics surfaces it before the LLM tries.)
• Missing MCP dependencies — same idea against disabled Mcp rows.
• Unknown ACPX agents — for skills with runtime: acpx, flags any acpx_agent value that isn't in the AcpAgent table (typo, removed CLI, etc.).
• Orphan DB rows — Skill rows whose SKILL.md file no longer exists on disk. Usually a sign that someone deleted a skill directory without running Reload.

Each section is collapsed when clean (✓ green) and expanded with red ⚠ counts when something's wrong. Run it after editing SKILL.md files or after toggling tools/MCPs to confirm nothing is silently broken. Backed by GET /agent/skills/_/diagnostics/ — pure read, no writes.

ACPX-Skills -> Reload Registry

A single-click button that re-runs the registry boot pipeline: rescan agent/skills_pkg/, refresh every Skill DB row's metadata (description, runtime, frontmatter_json, body_sha256), prune any DB row whose SKILL.md is gone. The user-toggled enabled field is preserved across reload.

Use this after you've authored or edited a SKILL.md on disk — no server restart needed. The success toast tells you the new skill count.

What the DB stores — and what it does NOT

By design, the Skill DB table stays at "enumeration + enable/disable" only, exactly the way the Tool and Mcp tables work. Per-skill permissions (filesystem read/write globs, allowed shell commands, network deny/allow), budgets (max_iterations / max_seconds / max_tokens), and the skill's body all live in the SKILL.md frontmatter on disk and are the only source of truth. The admin UI deliberately does NOT let you override them from the browser — if you want to change a permission, edit the SKILL.md and click Reload. This keeps git diff honest: every behavioural change to a skill shows up in a file, not in a database row that the next backup would silently archive.

Where to look

• HTTP endpoints: agent/views.py (list_skills_view, skill_detail_view, reload_skills_view, skills_diagnostics_view) — wired in agent/urls.py.
• WebSocket toggle: agent/consumers.py::receive (set-skills branch) → save_skill(name, enabled). The connect path also calls skill_establishment() for every row so the frontend's skills = [] cache hydrates on session start, mirroring how tools[] and agents[] hydrate.
• Tool-surface gating: agent/acpx/tools.py::_disabled_skill_names() — fails open on DB exception so a broken admin layer never silently hides skills.
• Frontend dialogs: agent/static/agent/js/skills_dialog.js (the Configure / Browse / Diagnostics / Reload dialogs) + agent/static/agent/css/skills_dialog.css.
• Coverage: 14 tests in agent/tests.py (SkillsAdminEndpointTests, SkillsToolSurfaceGatingTests, SkillsNavbarTemplateContractTests).

3.13. Tutorial: command a window from chat (chat_agent_windower)

Windower is the desktop window manager of the chat surface — the third member of the desktop-UI trio: where Mouser clicks inside a window and Keyboarder types into one, Windower commands the window itself. It is implemented self-contained on the Win32 API (pywin32 win32gui/win32con/win32process + ctypes), porting the window-management subset of Microsoft's Windows-MCP — including the cross-process AttachThreadInput focus-transfer dance that lets a background process reliably raise a foreground window. It is Windows-only and state-changing, so it appears in the Exec Report.

Tick only the Multi-Turn checkbox (Windower is a normal Multi-Turn tool — it is not behind the ACPX/Skill surface, so the ACPX checkbox is not required). Then ask, for example:

"Open Notepad, bring it to the front and maximize it, then tell me its size."

Tlamatini will launch the app (chat_agent_executer), confirm it is up (chat_agent_window_present), then call chat_agent_windower with action='maximize' and window_title='Notepad' — and you will watch the window come to the foreground and fill the screen. The tool promotes its result fields (action, window_title, matched, match_count, state, left, top, width, height) to the top level of its JSON return, so the answer can report the live geometry without parsing logs.

action can be any of: list (enumerate every open window with its geometry + state), focus, minimize, maximize, restore, move, resize, move_resize, close (by title), topmost / untopmost (always-on-top), or arrange (snap/tile to left/right/top/bottom halves, the four quadrants, center, or full). Matching is by match_mode ∈ substring (default) / exact / regex, with match_index to pick among same-titled windows. Use Windower — not Mouser — whenever the goal is the window as a whole (bring to front, tile, resize, close by title).

Two ready-made showcases live in the chat Prompts dropdown (the Catalog of Prompts): WINDOW SPOTLIGHT (basic — maximize + list + close) and WINDOW CHOREOGRAPHY (medium — restore → tile left → tile right → top-left quadrant → move/resize → list → close, so a single window visibly dances around the screen). Pick one, send, and watch.

3.14. Tutorial: drive a browser from chat (chat_agent_playwrighter)

Playwrighter drives a real browser (Playwright — Chromium / Firefox / WebKit) through a scripted, interactive, stateful flow. Where Crawler does a one-shot static fetch and the googler tool only searches, Playwrighter clicks, fills forms, waits for elements, extracts text/attributes, screenshots, asserts, and downloads — so it can log into a site, submit a multi-step form, click through a wizard, or scrape a JavaScript-rendered single-page-app behind a login. It needs Playwright installed (pip install playwright && playwright install); set headless=false to watch it drive, and hold_open_seconds=N (alias hold_open_ms) to keep the browser visible for N seconds after the last step before it closes — that's the "wait a few seconds before closing so I can see it" knob; just ask Tlamatini to wait and it passes it for you.

Tick only the Multi-Turn checkbox (Playwrighter, like Windower, is a normal Multi-Turn tool — ACPX is not required). Then ask, for example:

"Open Wikipedia in a visible browser, search for ‘Nahuatl’, and tell me the first paragraph of the article."

Tlamatini calls chat_agent_playwrighter with start_url, headless='false' (and hold_open_seconds if you asked it to keep the browser open), and the whole script as a single JSON string in steps_json (the flat key=value request grammar cannot express a list-of-dicts, so the script is passed as JSON and the agent json.loads it). Each step is {"action": <verb>, ...}; supported verbs are goto, click, dblclick, fill, type, press, select, check/uncheck, wait_for, wait, extract_text, extract_attr, screenshot, assert_visible, assert_text, and download. The run reports status, final_url, steps_run, assert_result, and the extracted values, so a downstream step (or a Forker, on the canvas) can branch on the verdict.

Two ready-made showcases live in the Prompts dropdown: BROWSER SPOTLIGHT (basic — open example.com with a visible browser, extract the heading, assert the link, screenshot) and BROWSER WIZARD (medium — a visible multi-step Wikipedia search: fill → click → wait → extract → assert → screenshot). The canvas counterpart is the visual Playwrighter node (see §4 and §9.5); the YAML steps list is its authoring form.

3.15. Tutorial: run Kali Linux tools from chat (chat_agent_kalier)

Kalier bridges Tlamatini to Kali Linux offensive-security tooling through the MCP-Kali-Server. That project runs a small Flask API server (server.py) on the Kali box exposing /api/command, /api/tools/<tool> and /health; Kalier talks to it directly over HTTP (Python-stdlib urllib, no extra packages in the agent pool), so it is the canonical tool for AI-assisted penetration testing, recon, and CTF solving. It is state-changing, so it appears in the Exec Report.

First, get the MCP-Kali-Server (server.py) running on your Kali machine and reachable from Tlamatini. Tlamatini is the embedded client — you no longer need Claude Desktop's client.py; instead you set the Kali box URL once in Config ▸ URLs → Kali server (Kalier) (the kali_server_url key in config.json, default http://127.0.0.1:5000). That default already works when Kali runs in WSL2 with localhost forwarding or when you SSH-tunnel the port (ssh -L 5000:localhost:5000 user@KALI_IP); for a LAN Kali box set it to http://<KALI_LAN_IP>:5000. (See Tlamatini-Kali-Setup.md for the full zero-client walkthrough.) ⚠️ Authorized targets only — run Kalier solely against systems you own or are explicitly authorized to test (engagement, lab, CTF).

Tick only the Multi-Turn checkbox (Kalier is a normal Multi-Turn tool — not behind the ACPX/Skill surface). Then ask, for example:

"Scan 10.0.0.5 with an nmap -sCV on ports 1-1000 and summarize the open services."

Tlamatini calls chat_agent_kalier with action='nmap', target='10.0.0.5', scan_type='-sCV', ports='1-1000' — and auto-injects your configured server_url from kali_server_url, so you never repeat the Kali box address in a prompt (the LLM only passes server_url= explicitly to hit a different one-off box). The action field selects the capability: command (any shell command on the Kali box), nmap, gobuster, dirb, nikto, sqlmap, metasploit, hydra, john, wpscan, enum4linux, or health (probe the server and which tools are installed — a good first call when you are unsure the API is reachable). The tool returns the Kali tool's stdout/stderr verbatim and captures an INI_SECTION_KALIER block (action, endpoint, subject, return_code, success, timed_out, server_url) for the Exec Report and Parametrizer.

On the canvas the same capability is the visual Kalier node (see §4 and §9.5): chain Starter → Kalier (nmap) → Parametrizer → Kalier (gobuster) → Forker → Ender to build a fully unattended, branch-on-result assessment pipeline. The visual node and the chat tool share the same MCP-Kali-Server contract.

3.16. Tutorial: build and flash STM32 firmware from chat (chat_agent_stm32er)

STM32er bridges Tlamatini to STM32 microcontroller firmware development through the STM32 Template Project MCP — a FastMCP server that exposes project scaffolding, build, flash, serial / SWD observation, and reset. It is state-changing (it compiles firmware and writes to hardware), so it appears in the Exec Report. The visual canvas counterpart is the STM32er node (see §4 and §9.5); both surfaces share the same MCP contract.

Zero-config auto-bootstrap — you only install STM32CubeIDE + Tlamatini. With no on-disk server_script configured (the new default), STM32er downloads the MCP from its git repo (a shallow git clone, or a GitHub-zip fallback when git is absent) into a per-user cache (%LOCALAPPDATA%/Tlamatini/STM32TemplateProjectMCP), pip-installs its deps (mcp + pyserial) if they are missing, and validates the server — all on first use. Nothing to clone by hand, no path to set. A bootstrap action triggers this explicitly; auto_bootstrap (default true) does it lazily before the first real action.

⚠️ Critical-mission safety preflight. Before it compiles or flashes anything, STM32er runs a validate preflight that checks the environment: arm-none-eabi-gcc, STM32CubeIDE, make/cmake, STM32_Programmer_CLI, the ST-LINK USB driver, a positively-confirmed connected ST-LINK probe, and the target device family. If the environment is wrong — or the request would target the wrong STM32 family — STM32er refuses rather than producing or flashing mis-targeted firmware. Hardware is conditional: compile-only actions (build, list_artifacts, clean, create_project, write_source) need no board; hardware actions (flash, erase, reset, serial_*, SWD reads, live_*) require a connected ST-LINK. The bundled MCP template is STM32F407VG-specific, so a cross-STM32F-family device mismatch is refused (a multi-family MCP fork is future work).

Tick only the Multi-Turn checkbox (STM32er is a normal Multi-Turn tool — not behind the ACPX/Skill surface). Then ask, for example:

"Scaffold a blinky project for the STM32F407, build it, and flash it to the connected board."

Tlamatini calls chat_agent_stm32er — bootstrapping the MCP if needed, running the safety preflight, then driving the build and flash. The action field selects one of 23 MCP tools (project scaffold / build / flash / erase / reset / serial / SWD / observe / …), 2 composites (serial_session, live_monitor), or 2 meta-actions (bootstrap, validate). The tool captures an INI_SECTION_STM32ER block for the Exec Report and Parametrizer.

On the canvas the same capability is the visual STM32er node: chain Starter → STM32er (validate) → Forker → STM32er (build) → STM32er (flash) → STM32er (serial_session) → Ender to build a fully unattended, validate-gated firmware pipeline. Three ready-made catalog demos ship in migration 0103: STM32 GENESIS (bootstrap + validate + compile, no board needed), STM32 BLINKY (validate + build + flash), and STM32 HIL OBSERVATORY (a validate-gated real-hardware flash + SWD + serial + reset). The zero-config end-to-end path (download → build → flash → reset) is verified on a real STM32F407G-DISC1, with 122 automated tests in agent/test_stm32er_agent.py.

📟 Serial output on a Discovery board needs a wire — the VCP is not bridged to the MCU's UART. This matters specifically for the 3rd STM32er demo, STM32 HIL OBSERVATORY, whose serial_session step reads the firmware's BOOT tlamatini hil count=… banner. The on-board ST-LINK on the STM32F4-Discovery family (including the STM32F407G-DISC1) provides debug (SWD) but — unlike ST Nucleo boards — does not internally route its USB Virtual COM Port to any of the target STM32F407's USART pins. A firmware that prints over USART2 (PA2 = TX, PA3 = RX) therefore shows nothing on the ST-LINK VCP regardless of baud or timeout: on the PCB those bytes have nowhere to go. To actually read that stream you must bridge the port yourself with an external USB-to-UART (USB-TTL) adapter — cross-wire adapter RX ← PA2, adapter TX → PA3, GND ↔ GND — and point serial_session at that adapter's COM port (not the ST-LINK VCP). No wiring is needed for the demo's primary hardware proof: the live_monitor step samples the live g_blink_count counter from the running MCU over the SWD debug channel, so it works on a bare, unwired board — which is exactly why STM32er treats an empty VCP read on a Discovery board as expected, not a failure, and why the HIL demo proves the firmware is alive over SWD first and treats the serial banner as a bonus.

3.17. Tutorial: build and upload ESP32 firmware from chat (chat_agent_esp32er)

ESP32er bridges Tlamatini to ESP32 (and ESP8266 / Espressif) firmware development through PlatformIO Core — the open-source embedded build system whose pio CLI covers build, upload, serial monitor, debug, the board database, the package manager and static analysis. Unlike STM32er — which drives a separate MCP server because STM32CubeIDE has no unified CLI — PlatformIO already ships a complete CLI, so ESP32er invokes pio subcommands directly (no MCP server). It is state-changing (it compiles firmware and writes to hardware), so it appears in the Exec Report. The visual canvas counterpart is the ESP32er node (see §4 and §9.5).

Zero-config auto-bootstrap — you only install the board USB driver + Tlamatini. With no on-disk pio_executable configured (the default), ESP32er downloads PlatformIO Core itself — the official get-platformio.py installer (with a pip install platformio fallback) into a per-user cache (%LOCALAPPDATA%/Tlamatini/platformio) — and validates it, all on first use. Nothing to install by hand. A bootstrap action triggers this explicitly; auto_bootstrap (default true) does it lazily before the first real action. (The first build then downloads the espressif32 platform + toolchain, hundreds of MB, so it is slow once.)

⚠️ Safety preflight. Before it builds or uploads anything, ESP32er runs a validate preflight: pio must be resolvable and a platformio.ini must exist, and for an upload/monitor a serial port must actually be connected (probed via pio device list). If the environment is wrong, ESP32er refuses rather than running a build/upload that cannot succeed. ESP32 flashes over its onboard USB-serial bootloader, so no external JTAG probe is needed to upload — only step-debug would need one. A non-espressif32 platform is a warning, not a refusal (PlatformIO is multi-target — there is no shared-linker-script risk like STM32 has).

Tick only the Multi-Turn checkbox (ESP32er is a normal Multi-Turn tool — not behind the ACPX/Skill surface). Then ask, for example:

"Scaffold a blinky project for an esp32dev board at C:/esp/blink, build it, and flash it to the connected board, then show me the serial output."

Tlamatini calls chat_agent_esp32er — bootstrapping PlatformIO if needed, running the safety preflight, then driving create_project → write_source → build → upload → monitor. The action field selects the capability: environment/meta (bootstrap, validate, system_info, boards); project lifecycle (create_project, write_source, read_source, list_sources, clean); build & flash (build, upload, build_and_upload, list_artifacts); serial HIL (device_list, monitor, monitor_session); packages & QA (pkg_install, pkg_list, pkg_update, check, test). The tool captures an INI_SECTION_ESP32ER block for the Exec Report and Parametrizer. Headless step-debug (pio debug) is a planned fast-follow (it needs JTAG — an external probe or the built-in USB-JTAG on ESP32-S3/C3).

On the canvas the same capability is the visual ESP32er node: chain Starter → ESP32er (create_project) → Parametrizer → ESP32er (write_source) → ESP32er (build) → ESP32er (upload) → ESP32er (monitor) → Ender to build a fully unattended firmware pipeline. The agent is implemented self-contained with the Python stdlib (subprocess + urllib), with automated tests in agent/test_esp32er_agent.py.

3.18. Tutorial: build and upload Arduino firmware from chat (chat_agent_arduiner)

Arduiner bridges Tlamatini to Arduino firmware development through the Arduino CLI — Arduino's official command-line tool for cores, libraries, compile, upload and serial monitor. It is the third microcontroller-firmware agent and the direct-CLI sibling of ESP32er: like PlatformIO's pio (and unlike STM32er's MCP server), arduino-cli is itself a complete CLI, so Arduiner invokes arduino-cli subcommands directly (no MCP server). It is state-changing (it compiles firmware and writes to hardware), so it appears in the Exec Report. The visual canvas counterpart is the Arduiner node (see §4 and §9.5).

The microcontroller is selected by fqbn. The board's Fully Qualified Board Name — e.g. arduino:avr:uno, arduino:avr:mega2560, arduino:samd:mkr1000, esp32:esp32:esp32 — chooses the target chip; port + baud set the upload/monitor link.

Zero-config auto-bootstrap — you only install the board USB driver + Tlamatini. With no on-disk arduino_cli_executable configured (the default), Arduiner downloads the arduino-cli binary itself — the platform release archive from downloads.arduino.cc, unzipped into a per-user cache (%LOCALAPPDATA%/Tlamatini/arduino-cli) — then runs config init + core update-index, all on first use. (arduino-cli is a Go binary, so the bootstrap is a binary download, not a pip install.) Auto-core-install: unlike PlatformIO, arduino-cli does not auto-install platforms on compile, so before a build/upload Arduiner derives the FQBN's platform and, when it is missing, runs core update-index + core install (honoring additional_urls for third-party ESP32 / STM32 / RP2040 cores). A bootstrap action triggers the binary download explicitly; auto_bootstrap and auto_core_install (both default true) do it lazily.

⚠️ Safety preflight. Before it builds or uploads anything, Arduiner runs a validate preflight: arduino-cli must be resolvable, a sketch (.ino) and an FQBN must exist, and for an upload/monitor a serial port must actually be connected (probed via arduino-cli board list). If the environment is wrong, Arduiner refuses rather than running a build/upload that cannot succeed. A malformed FQBN is a warning, not a refusal (arduino-cli is multi-vendor — no shared-linker-script risk like STM32 has).

Tick only the Multi-Turn checkbox (Arduiner is a normal Multi-Turn tool — not behind the ACPX/Skill surface). Then ask, for example:

"Create a blinky sketch for an Arduino Uno at C:/ino/blink, build it, and upload it to the connected board, then show me the serial output."

Tlamatini calls chat_agent_arduiner — bootstrapping the CLI and installing the FQBN's core if needed, running the safety preflight, then driving create_project → write_source → build → upload → monitor. The action field selects the capability: environment/meta (bootstrap, validate, system_info, boards, device_list); cores & libraries (core_update_index, core_search, core_list, core_install, core_uninstall, lib_update_index, lib_search, lib_list, lib_install); project lifecycle (create_project, write_source, read_source, list_sources); build & flash (build, upload, build_and_upload, clean, list_artifacts); serial HIL (monitor, monitor_session). create_project scaffolds from the bundled ArduinoTemplateProject (the Arduino analog of STM32er's STM32 Template Project and ESP32er's pio scaffold) and stamps the FQBN/port into the template's sketch.yaml profile. The tool captures an INI_SECTION_ARDUINER block for the Exec Report and Parametrizer.

On the canvas the same capability is the visual Arduiner node: chain Starter → Arduiner (create_project) → Parametrizer → Arduiner (write_source) → Arduiner (build) → Arduiner (upload) → Arduiner (monitor) → Ender to build a fully unattended firmware pipeline. The agent is implemented self-contained with the Python stdlib (subprocess + urllib + zipfile/tarfile), with automated tests in agent/test_arduiner_agent.py. (Note: the first core install + compile downloads the board toolchain, so it is slow once.)

4. Visual Workflow Designer (/agentic_control_panel/)

The chat is great for one-off tasks. The designer is for jobs you want scheduled, unattended, or identically reproducible.

4.1. Canvas anatomy

┌────────────────────────────────────────────────────────────────────────┐
│ ▶ Start  ⏸ Pause  ⏹ Stop  ⚠ Hypervisor  💾 Save  📂 Load  ✓ Validate │
├──────────────────┬─────────────────────────────────────────────────────┤
│ Sidebar          │                                                     │
│ ─ Control        │                                                     │
│   Starter, Ender │            CANVAS (#canvas-content)                 │
│ ─ Routing        │       (draggable agents, typed connections,         │
│   Forker, Asker  │        green-running / red-down / yellow-paused     │
│ ─ Logic Gates    │        LEDs)                                        │
│   AND OR Barrier │                                                     │
│ ─ Action / etc.  │                                                     │
└──────────────────┴─────────────────────────────────────────────────────┘

• The canvas scrolls: viewport is #submonitor-container, content layer is #canvas-content. New canvas-level features should be children of #canvas-content.
• Connections are typed: green = "start the target after this finishes" (target_agents), blue = "monitor this source's log" (source_agents).
• Double-click an agent to edit its config. Right-click for description / log / explore-dir / open-cmd / restart.

4.2. Tutorial: your first flow (3 agents)

Goal: run a shell command, take a screenshot, end.

1. Drag Starter onto the canvas (top-left).
2. Drag Executer to its right.
3. Drag Shoter further right.
4. Drag Ender at the far right.
5. Connect: Starter → Executer → Shoter → Ender (drag from the right edge of one to the left edge of the next).
6. Double-click Executer, set command to dir C:\ (or ls /tmp).
7. Double-click Shoter, set output_dir to a writable folder.
8. Leave Ender wiring to Tlamatini. Validate/Start will calculate Ender's target_agents kill list from the arrows.
9. Click ✓ Validate — Tlamatini compiles the visible canvas, then runs structural checks (no orphans, no self-connections, terminal agents reachable).
10. Click ▶ Start. LEDs go green, then gray. Open output_dir — there's a screenshot.

4.3. Saving and loading .flw files

💾 Save — pick a name. You get a JSON file with positions, configs, and connections. Distribute to colleagues; they 📂 Load the same file and run the same flow. .flw is also what the chat's Create Flow button emits.

A .flw file is meant to describe the idea of the flow, not the exact machine it was created on. A good .flw says:

• "There is a Starter here."
• "There is an Executer there."
• "Starter connects to Executer."
• "Executer uses this script."

It should not say:

• "This flow only works from C:/Development/Tlamatini/...."
• "This flow only works from the install folder on Angel's PC."
• "This Parametrizer mapping exists somewhere in a temporary pool directory, good luck."

Saved flows now carry a small schemaVersion plus an artifacts section. The most important artifact today is Parametrizer mappings. When you save a flow with a Parametrizer, Tlamatini keeps the mapping in the .flw. When you load the flow later, Tlamatini recreates interconnection-scheme.csv for that Parametrizer in the current session pool.

For a beginner, the practical rule is simple: if you configured Parametrizer with the mapping dialog, Save/Load should remember that mapping.

4.4. Validate and Start now compile the live canvas

This is the most important reliability change in the visual designer.

Before, Validate mostly read whatever agent configs already existed in the pool directory. That could become stale:

1. You drag nodes around.
2. You load a .flw.
3. You edit a config.
4. You reconnect an edge.
5. The pool directory still contains an older config.yaml.
6. Validate or Start reads that older file and acts confused.

Now ACP takes a fresh snapshot of the canvas before validation and start. The snapshot includes:

• every visible node;
• each node's position;
• every connection;
• input and output slot numbers;
• current in-browser config;
• Parametrizer mappings.

The backend then compiles that snapshot into real pool config.yaml files using the Agent Contract Registry. In beginner terms: the picture on the screen becomes the source of truth.

Another important nuance: if you opened an agent dialog and manually edited wiring-sensitive fields such as source_agents or an Ender kill list, those dialog edits now survive compilation. Canvas edges still contribute their live connections, but a deliberate dialog override is no longer silently discarded by Validate or Start.

What happens when you click ✓ Validate:

1. Browser captures the live canvas.
2. Backend compiles it in dry-run mode.
3. Frontend validates the compiled configs.
4. Nothing is written to disk just for validation.

What happens when you click ▶ Start:

1. Browser captures the live canvas.
2. Backend compiles it in write mode.
3. Pool folders/configs are updated.
4. Logs are cleared.
5. Starter agents launch.

This removes a whole class of "I swear I connected it correctly, why is it running the old thing?" problems.

4.5. Pause, Resume, Stop

This is why long-running workflows (Crawler scraping 10k URLs, Parametrizer iterating segments) survive pauses.

Stop is also safer in mixed flows now: the ACP cleanup path is better at terminating leftover session processes before the next run begins, so partially mixed ACP sessions are less likely to leave orphaned agents behind.

4.6. FlowHypervisor (watchdog)

Click ⚠ Hypervisor — a system FlowHypervisor agent starts watching every running agent. It is an LLM that reads each agent's log, builds an NxN connection matrix from the canvas wiring, and emits exactly OK or ATTENTION NEEDED { explanation }. If it raises, the browser pops an alert. Add custom rules to user_instructions in its config.

4.7. FlowCreator (let an LLM design the flow)

Drag FlowCreator, double-click, and type a natural-language objective:

"Every hour, crawl our status page; if it shows ERROR, email the on-call engineer."

Click Generate. FlowCreator reads agentic_skill.md (its design playbook), produces a JSON flow description, and renders agents + connections onto the canvas. Tweak and run.

4.8. Parametrizer (chain outputs into the next agent's config)

Tlamatini agents communicate through log files and config.yaml. Parametrizer is the bridge: it reads structured segments from a source agent's log, injects mapped values into a target agent's config.yaml, runs the target, restores the config, advances the cursor, repeats.

The unified output format every Parametrizer-friendly agent emits:

INI_SECTION_<AGENT_TYPE><<<
key1: value1
key2: value2

multi-line body content (becomes 'response_body')
>>>END_SECTION_<AGENT_TYPE>

35 source agents support this format: Apirer, Gitter, Kuberneter, Crawler, Summarizer, File-Interpreter, Image-Interpreter, File-Extractor, Prompter, FlowCreator, Kyber-KeyGen/Cipher/DeCipher, Gatewayer, Gateway-Relayer, De-Compresser, Googler, Playwrighter, ACPXer, Shoter, Camcorder, Recorder, Whisperer, AudioPlayer, VideoPlayer, Talker, Mouser, Windower, Unrealer, Reviewer, Analyzer, Kalier, STM32er, ESP32er, Arduiner.

Canonical example:

Apirer ─► Parametrizer ─► Kyber-Cipher

Apirer hits 3 endpoints → 3 INI_SECTION_APIRER<<< blocks → Parametrizer maps response_body → buffer → Kyber-Cipher runs 3 times, encrypting each body. No manual config editing. Pause-safe. Single-lane queue.

The mapping dialog is now part of normal flow persistence:

1. Connect exactly one source into Parametrizer.
2. Connect Parametrizer to exactly one target.
3. Double-click Parametrizer.
4. Click a source field on the left.
5. Click a target config field or target marker on the right.
6. Save mappings.
7. Save the .flw.

When the .flw is loaded later, Tlamatini restores the mappings and writes the Parametrizer's interconnection-scheme.csv again. You do not need to remember which pool directory had the CSV.

One limitation is intentional: one Parametrizer is a single-lane queue from one source to one target. If one API response must feed Emailer and File-Creator, use two Parametrizers.

4.9. Gatewayer (external triggers)

Two trigger modes:

Pending events survive crashes via reanim_queue.json. To accept GitHub-style webhooks (which sign only the body), put the bundled Gateway-Relayer in front.

5. ACPX — External Coding-Agent CLIs as Tools

5.1. What ACPX is

ACPX = Agent Communication Protocol eXtension. It spawns external coding-agent CLIs as out-of-process child processes, talks to them over stdin/stdout, persists the conversation as NDJSON transcripts, and brokers them to the chat LLM as 12 native tools. It is a Python port of OpenClaw's ACPX plugin — agent_id mapping, permissionMode vocabulary, and SKILL.md frontmatter all match verbatim.

5.2. Supported agent_ids and transports

Defined in agent/acpx/agent_registry.py::DEFAULT_ACP_AGENTS. User overrides go in config.json under acpx.agents.<id>.

Transport modes:

• oneshot-prompt — fresh process per turn; prompt is a CLI arg; stdin closes; stdout captured to EOF. The only transport that reliably captures TUI agents on Windows (TUI CLIs detect a piped stdout and refuse to flush in long-lived mode).
• json-acp — child speaks one JSON envelope per turn, ends with {"done": true}.
• tui-repl — long-lived REPL; transport-aware idle rule fires after startup_grace + idle_seconds even with zero events (a silent TUI is, by definition, finished).

5.3. The 12 ACPX/Skill tools

All return JSON envelopes. Failures: {"ok": false, "reason": "...", "code": "..."}.

27 seed skills live under agent/skills_pkg/ (acp-router, summarize, setup-new-acpx-key, skill-creator, flow-making, code-review, security-audit, kali-pentest, the create-new-agent / create-new-mcp authoring runbooks, hello-world, 8× tlamatini-* maintenance helpers incl. flow-from-objective → which now delegates to flow-making, plus OpenClaw-format ports for github / gmail / slack / jira / notion / todoist / trello / weather).

5.4. Tutorial: spawn-and-go (single agent)

Tick Multi-Turn + ACPX + Exec Report and send:

"Spawn claude in C:/Development/Tlamatini, ask it to list the top-level files, harvest the answer, and kill the session."

Expected tool sequence:

acp_doctor
  → acp_spawn(agent_id="claude", task="list top-level files")
    → acp_send_and_wait(session_id, "...")
      → acp_kill(session_id)

5.5. Tutorial: multi-CLI relay with acp_relay

"Spawn claude in this dir, ask it to draft a refactor of worker.py. Spawn gemini, relay claude's answer to it, ask gemini to critique. Kill both."

Expected sequence:

acp_doctor
  → acp_spawn(claude, draft_task)
    → acp_send_and_wait(session_a, …)
      → acp_spawn(gemini, critique_template)
        → acp_relay(session_a, session_b)     # ONE call — transform=last_assistant_text
          → acp_kill(session_a)
            → acp_kill(session_b)

Without acp_relay, that hand-off is three calls (acp_transcript → string-manipulate → acp_send). Always prefer the dedicated tool.

5.6. API key setup (the easy button)

Two layers exist in config.json:

{
  "ANTHROPIC_API_KEY": "sk-ant-...",     // Layer 1: Tlamatini's own cloud calls
  "GEMINI_API_KEY": "AIza...",
  "acpx": {
    "agents": {
      "claude": { "env": { "ANTHROPIC_API_KEY": "sk-ant-..." } },   // Layer 2: spawned child env
      "gemini": { "env": { "GEMINI_API_KEY": "AIza...", "GOOGLE_API_KEY": "AIza..." } },
      "codex":  { "env": { "OPENAI_API_KEY":  "sk-..." } },
      "qwen":   { "env": { "DASHSCOPE_API_KEY": "sk-..." } }
    }
  }
}

Merge order at spawn: {**os.environ, **spec.env} — explicit per-agent env wins over an exported shell variable.

Easier path — invoke the setup-new-acpx-key skill from chat (Multi-Turn + ACPX ticked):

"Use invoke_skill with setup-new-acpx-key to register my Anthropic key for the claude agent_id." (paste the key)

The skill writes data.keys, patches both config.json layers, optionally extends regen_secrets.py, and verifies via acp_doctor.

Security: config.json is git-tracked. Use python regen_secrets.py --mode push-able to swap real keys for placeholders before commit; --mode keyed restores from data.keys (gitignored). Never commit data.keys.

5.7. ACPXer — the visual canvas counterpart

ACPXer is the canvas-facing version of the 12 LLM-facing tools. One ACPXer node = one full ACPX session lifecycle. It is self-contained — does NOT import agent.acpx — because pool subprocesses can't import agent.*. Mirrors the runtime's transport-aware drain inline (~120 lines), writes byte-identical NDJSON transcripts, and emits Parametrizer-compatible INI_SECTION_ACPXER<<< blocks.

Canonical visual relay flow:

Starter → ACPXer(claude) → Parametrizer → ACPXer(gemini) → Parametrizer → ACPXer(cursor) → File-Creator → Ender

Three different LLMs argue back and forth, fully visual, fully unattended.

6. Unreal MCP — Driving Unreal Engine 5 from Tlamatini

The Unrealer agent (#62 in the catalog) lets Tlamatini drive a live Unreal Engine 5 editor through the Unreal MCP plugin's TCP socket protocol. You spawn a chat_agent_unrealer call from Multi-Turn or drop an Unrealer node on the visual canvas; Tlamatini opens a TCP connection to 127.0.0.1:55557, sends one JSON command ({"type": <verb>, "params": {...}}), captures the engine's JSON response into an INI_SECTION_UNREALER<<< block, and triggers downstream agents. Because the agent forwards whatever command + params you give it, the catalog is exactly whatever your connected plugin build exposes — from the base 28-command upstream release up to the 53-command, nine-category extended surface (actor manipulation incl. viewport screenshots, Blueprint creation and graph wiring, input mappings, UMG widget building, in-editor Python/console execution, level I/O, asset import, and material authoring) shipped by Tlamatini's own plugin fork, XAIHT/XaihtUnrealEngineMCP (the Unreal Engine MCP modified specifically for this system — see §6.2) — without you ever leaving the chat or the canvas.

6.1. What Unreal MCP is

Unreal MCP is an open-source UE5 plugin (Model Context Protocol over a TCP socket) that runs inside the Unreal Editor process and accepts one JSON command per TCP connection. Each command names a verb (spawn_actor, create_blueprint, compile_blueprint, add_widget_to_viewport, …) and a dictionary of parameters; the plugin schedules the work onto UE5's game thread, executes it, and writes a JSON response back over the same socket before closing it. The wire shape is small — {"type": <command>, "params": {...}} going in, {"status": "ok"|"error", "result": {...}} (or {"success": false, "error": "..."}) coming back — and is identical across every documented command.

Tlamatini does not embed or compile the plugin. It is a client of whatever UE5 instance the user has already started. The engine must be open, the plugin must be enabled, and its in-engine listener must be bound to 127.0.0.1:55557 (the default — configurable per-call via host / port). Tlamatini contributes the calling side: one wrapped Multi-Turn tool, one visual canvas node, one Agent Contract entry, one Exec Report row family, and one Parametrizer source mapping — all built around the same UnrealConnection adapter at agent/agents/unrealer/unrealer.py.

6.2. The MCP plugin source (the MCP git location)

Recommended — Tlamatini's own extended fork. The plugin Tlamatini is built and tested against is the Unreal Engine MCP modified specifically for this system:

• Repository: https://github.com/XAIHT/XaihtUnrealEngineMCP.git
• What it is: the canonical chongdashu/unreal-mcp plugin forked and extended for Tlamatini. It ships the full 53-command, nine-category surface this chapter documents — the base editor / blueprint / node / project / umg verbs plus the System / Level / Asset / Material families and the newer take_screenshot / focus_viewport / set_pawn_properties / find_blueprint_nodes verbs.
• Plugin folder name: UnrealMCP
• Default plugin TCP port: 55557 on 127.0.0.1
• Supported UE versions: Unreal Engine 5.5+

It speaks the identical wire protocol on the identical port as every other build below, so it is a drop-in: Tlamatini's UnrealConnection adapter needs no client-side changes to use it. Install this one if you want the System / Level / Asset / Material families that the seeded demos idPrompt 60/61/62 (§6.5) exercise.

Upstream base. The XAIHT fork is built on the canonical reference implementation Tlamatini's UnrealConnection adapter mirrors verbatim:

• Repository: https://github.com/chongdashu/unreal-mcp
• License: MIT
• Supported UE versions: Unreal Engine 5.5+

If you only ever need the base 28-command surface (editor / blueprint / node / project / umg), the upstream is enough on its own.

Equivalent community forks. Two other forks ship the same wire protocol on the same port; either works with Tlamatini's Unrealer with no client changes:

• https://github.com/CrispyW0nton/Unreal-MCP-Ghost
• https://github.com/gingerol/vhcilab-unreal-engine-mcp

Pick the build that matches your UE5 version and your team's licensing comfort. If you fork the plugin to add a new command verb, your fork is automatically usable from Tlamatini — there is no client-side allow-list of verbs (the wrapped tool forwards any command + params pair verbatim).

6.3. Installing and enabling the plugin inside your UE5 project

The plugin is a per-project install (not engine-wide). Steps:

1. Clone or download the plugin — the recommended XAIHT/XaihtUnrealEngineMCP fork from §6.2, or any compatible build (only the MCPGameProject/Plugins/UnrealMCP folder matters — different forks may name the folder slightly differently; rename to UnrealMCP if needed).
2. Drop the folder into your project's Plugins/ directory so the final path is <YourProject>/Plugins/UnrealMCP/UnrealMCP.uplugin. Create the Plugins/ folder at the project root if it does not exist.
3. Open the project in UE5. The editor will detect the new plugin and offer to rebuild it for your engine version — accept. If you opened a Blueprint-only project, you will be prompted to install Visual Studio Build Tools / Xcode command-line tools first, since the plugin is C++.
4. Enable the plugin via Edit → Plugins → search "UnrealMCP" → tick Enabled. Restart the editor.
5. Confirm the listener is bound. With the editor running, open the Output Log (Window → Developer Tools → Output Log) and look for a line such as LogTemp: UnrealMCP listening on 127.0.0.1:55557. That line is your green light: the plugin is now waiting for JSON commands on the loopback interface.

You do not need to press Play (PIE). The plugin listens at editor level — actor manipulation, Blueprint creation, widget construction, etc. all work against the open project even when PIE is stopped. Some UMG operations (add_widget_to_viewport) physically render only after the user enters PIE, but the build steps are queued correctly either way.

6.4. The command catalog (up to 53 commands across 9 categories)

The Unrealer agent forwards whatever command + params you pass it, so the exact catalog is whatever your connected plugin build exposes — there is no client-side allow-list of verbs. The canonical chongdashu/unreal-mcp release ships 28 commands across 5 categories (rows marked base below); plugin builds that add the System / Level / Asset / Material command handlers — such as Tlamatini's own extended fork XAIHT/XaihtUnrealEngineMCP (§6.2) — bring the total to 53 commands across 9 categories:

execute_python is the universal escape hatch — it runs an arbitrary Python script inside the editor, so anything in UE5's unreal Python API (Niagara, Sequencer, landscape, audio, …) is reachable even when no dedicated verb exists. take_screenshot closes the observe→act loop: spawn or change something, then capture the viewport to verify it. Note that the plugin's headless build/cook/test tools (build_project, run_automation_tests, run_macro) are not part of this catalog — they shell out to UnrealEditor-Cmd as separate processes and are unreachable over the editor's TCP socket. Chain Unrealer nodes through a Parametrizer for the run_macro equivalent.

Param shapes vary per command (e.g. spawn_actor wants name + type + location + rotation; create_blueprint wants name + parent_class; set_material_parameter wants material + parameter + value; import_asset wants source_file (a disk path) + destination_path (a /Game content path)). The Unrealer agent does not validate them — it forwards them as-is, after two defensive fixups: it normalizes /Content/... content paths to /Game/..., prunes unset placeholder params, and remaps params.console_command → the wire's params.command for execute_console_command (so the console line doesn't collide with the top-level command: selector). The plugin will reply with {"status": "error", "error": "<reason>"} if a param is missing or malformed, and that error lands verbatim in the INI_SECTION_UNREALER block so Multi-Turn / Parametrizer can branch on it.

6.5. Using Unreal MCP from the chat (chat_agent_unrealer)

The wrapped Multi-Turn tool chat_agent_unrealer is the easiest way in. Tick Multi-Turn in the toolbar, leave Exec Report ticked too (the Unreal calls get their own table in the answer), and send a prompt like:

"Run Unreal command with command='spawn_actor' and params.name='MyCube' and params.type='StaticMeshActor' and params.location=[0,0,150]."

The planner picks chat_agent_unrealer, the wrapped runtime spawns one short-lived unrealer.py child under agent/agents/pools/_chat_runs_/unrealer_<seq>_<id>/, the child opens a TCP socket to 127.0.0.1:55557, sends the JSON command, captures the response, emits the INI_SECTION_UNREALER block to its log, and exits. The Multi-Turn loop reads the run's log excerpt, parses the section, and returns the full Unreal response JSON to the LLM. The LLM then sees the engine's reply and either reports it to you, branches on it, or fires the next call.

The tool accepts the same overrides documented in config.yaml:

• host='10.0.0.5' and port=55557 to target a remote UE instance (rare; the plugin binds to loopback by default and you would need to change the bind address inside the plugin or tunnel it).
• connect_timeout=5 and read_timeout=10 to widen the budgets for slow operations (e.g. compile_blueprint on a complex graph).

Built-in demo prompts. Migration 0087_add_unrealer_demo_prompt.py seeds a one-click demo into the Prompts table (idPrompt=25). Open the chat, click the Prompts dropdown, pick Unreal MCP End-to-End Editor Drive, and Tlamatini will execute ten guided steps spanning the base command categories — sanity-probe (get_actors_in_level), spawn a StaticMeshActor (spawn_actor), verify it (find_actors_by_name), scaffold a Blueprint (create_blueprint), add a StaticMeshComponent (add_component_to_blueprint), compile (compile_blueprint), spawn an instance (spawn_blueprint_actor), build a UMG HUD widget (create_umg_widget_blueprint → add_text_block_to_widget → add_button_to_widget → add_widget_to_viewport) — and render the result as a per-step HTML report table at the bottom of the answer. Use it as your smoke test the first time you wire the plugin up.

Migration 0100_add_unrealer_extended_demo_prompts.py adds three more demos that exercise the extended (System / Level / Asset / Material) surface the base demo never touches, at basic → hard complexity:

• idPrompt=60 — Unreal Snapshot (basic): the observe→act loop — get_current_level → spawn_actor → take_screenshot (to C:/Temp/unreal_snapshot.png) → save_current_level.
• idPrompt=61 — Unreal Scene Forge (medium): content authoring — list_assets → create_folder → create_material → create_material_instance → set_material_parameter → spawn_actor → assign_material → take_screenshot → save_all. (It is honest that set_material_parameter on a freshly-created blank material may return status: error — that is expected and recorded, not aborted.)
• idPrompt=62 — Unreal Python & Introspection (hard): the System escape hatch — execute_console_command (via the agent's params.console_command remap) → get_class_info → list_assets → execute_python (a multi-line script passed as a triple-quoted params.code) → take_screenshot.

All three drive chat_agent_unrealer exactly like the base demo (tick only Multi-Turn; ACPX not required) and require the same running editor + bound plugin listener.

6.6. Using Unreal MCP on the canvas (the visual Unrealer node)

For unattended .flw workflows, drop the Unrealer sidebar agent onto the canvas. Each node sends exactly one Unreal command when its turn arrives. The node's config.yaml is the same one shipped under agent/agents/unrealer/config.yaml:

host: 127.0.0.1
port: 55557
command: get_actors_in_level
params:
  name: ''
  type: ''
  location: []
  # ... (the shipped config carries empty placeholders for every param across
  #     all 9 categories — editor/blueprint/node/umg/system/level/asset/material
  #     — so the Flow Compiler's dotted `params.X` overrides always resolve into
  #     an existing YAML leaf. Unset placeholders are pruned before the command
  #     is sent, so add/remove keys freely to match the verb you picked.)
connect_timeout: 5
read_timeout: 10
source_agents: []
target_agents: []

The agent emits an INI_SECTION_UNREALER<<< block to its log, which means Parametrizer can chain Unreal calls together. Registered source fields (agent/services/agent_contracts.py): host, port, command, status, error, response_body. Canonical multi-step canvas pattern — create a Blueprint, compile it, spawn an instance:

Starter → Unrealer(create_blueprint) → Parametrizer → Unrealer(compile_blueprint)
       → Parametrizer → Unrealer(spawn_blueprint_actor) → Ender

Each Parametrizer copies the previous Unrealer's response_body (or a specific JSON field within it, via the Parametrizer dialog's interconnection-mapping UI) into the next Unrealer's params block. Branching on status (ok vs error) via a Raiser between Unrealer and the next Parametrizer gives you per-step exception handling — e.g., abort to a Notifier if compile_blueprint returns status: error.

6.7. What the agent actually does, end-to-end

The unrealer.py script (~120 lines of business logic, plus the standard pool-agent boilerplate) is self-contained: it does NOT import from agent.acpx, agent.services, or any other Tlamatini-internal package. Pool subprocesses run as separate Python interpreters with no sys.path back into the Django app, so the inline UnrealConnection adapter is a verbatim mirror of the upstream Unreal MCP Python client (with the FastMCP plumbing stripped out). Per execution:

1. Load config.yaml, read host, port, command, params, timeouts, and target_agents.
2. Write agent.pid so the orphan-process reaper (chapter 11) can track the run.
3. Open a fresh TCP socket to host:port with TCP_NODELAY, SO_KEEPALIVE, and 64 KB send/recv buffers. The Unreal MCP plugin closes the socket after each command, so the agent opens a new one per turn.
4. Send json.dumps({"type": command, "params": params}) (no trailing newline) and call recv() until a complete JSON document has been assembled (validated by attempting json.loads() on the accumulated bytes after each chunk).
5. Normalize the response shape — {"success": false, ...} from older plugin builds is rewritten into {"status": "error", "error": ...} so downstream Parametrizer / Multi-Turn code can rely on a single shape.
6. Emit one atomic logging.info() call with the INI_SECTION_UNREALER<<< block (header: host, port, command, status, error; body: pretty-printed Unreal JSON response, capped at 64 KiB). Single-call emission is mandatory for the parser at agent/agents/parametrizer/parametrizer.py.
7. Trigger target_agents even on error, so the canvas can route on the status field instead of relying on agent-level fail-stops.
8. Remove agent.pid and exit.

Failure modes the adapter handles gracefully (each turns into {"status": "error", "error": "<reason>"} plus a non-fatal log line, never an uncaught exception):

• Connection refused — the plugin's TCP listener is not bound (editor not running, plugin not enabled, port mismatched).
• Socket timeout during receive — UE5's game thread is busy (e.g. compile_blueprint on a heavy graph) and exceeded read_timeout. Raise read_timeout in config.yaml or in the wrapped tool call.
• Malformed JSON — the plugin closed mid-write; logged as an error status and downstream agents still fire.

6.8. Exec Report integration

chat_agent_unrealer is registered in _EXEC_REPORT_TOOLS (agent/mcp_agent.py) under agent_key="unrealer" and agent_display="Unrealer". When Exec Report is ticked alongside Multi-Turn, every Unreal call shows up as one row in a dedicated List of Unrealer Operations table at the bottom of the answer. Columns: command (left-bordered with the Unrealer caption gradient), success (SUCCESS / FAILURE derived from the underlying tool-call verdict — the same verdict Multi-Turn already uses for dedup and tool_calls_log). The table styling lives in agent/static/agent/css/agent_page.css (caption gradient mirrors .canvas-item.unrealer-agent in agentic_control_panel.css).

6.9. Bullet-proof checklist for Unreal Engine users

A short pre-flight you can copy into a sticky note before any session:

Then run the seeded Unreal MCP End-to-End Editor Drive demo prompt (idPrompt 25) as your smoke test. A clean run leaves three artifacts in your project: actor TlamatiniProbe_Cube, Blueprint BP_TlamatiniProbe with one spawned instance TlamatiniProbe_Spawned, and widget /Game/UI/WBP_TlamatiniProbeHUD. Delete them via right-click in the Content Browser when you are done.

6.10. Troubleshooting Unreal MCP

• acp_doctor is not relevant here. Unreal MCP is a workflow-agent surface, not the ACPX surface — the acp_* tools talk to external coding-agent CLIs (claude, gemini, …), not to UE5. The corresponding "is the channel alive?" probe for Unrealer is to call chat_agent_unrealer with command='get_actors_in_level' and check that status == 'ok'.
• status: error / Failed to connect to Unreal at 127.0.0.1:55557. The plugin is not listening. Check the UE5 Output Log for the UnrealMCP listening on … line. If the line is absent, the plugin is either disabled or failed to build (re-open the project; UE5 will re-prompt to rebuild). If the line is present but the connection still fails, your firewall is blocking loopback (rare on Windows, but Restart-Service mpssvc and re-test if you have aggressive endpoint security).
• Timeout receiving Unreal response. UE5's game thread is busy. Either widen read_timeout (config.yaml or the wrapped-tool call), or split the work into smaller commands (e.g. spawn 10 actors with 10 separate calls instead of one spawn_n_actors macro the plugin does not actually expose).
• status: error from a Blueprint command, but the verb seems valid. Capitalize the parent_class exactly as UE5 expects (Actor, Pawn, Character, UserWidget, …). The plugin does not auto-resolve actor → Actor.
• Widget appears in the Content Browser but is invisible in PIE. add_widget_to_viewport queues the widget at editor level; you still need to press Play in the editor (or call add_widget_to_viewport from within a running PIE session) to make it render. This is an Unreal MCP plugin behavior, not a Tlamatini bug.
• The Output Log shows the plugin received the command but nothing happened in the level. Most often: an actor spawn at a coordinate inside another object's collision volume. UE5 silently fails the spawn. Raise params.location by [0, 0, 150] and retry.
• A second instance of UE5 is bound to the same port. Only one UnrealMCP listener can bind to 127.0.0.1:55557 per host. Close the second editor instance, or configure each instance to bind to a different port and pass port=<n> per Unrealer call.

For the full debugging trail: pool-agent log lives at <pool>/unrealer_<n>/unrealer_<n>.log; chat-wrapped runs land under agent/agents/pools/_chat_runs_/unrealer_<seq>_<id>/unrealer_<seq>_<id>.log. Both contain the outbound JSON command and the inbound Unreal response verbatim.

7. Building a Frozen Distribution

For shipping a one-click Windows installer to end users.

7.1. Three-step pipeline

build.py  ──►  build_uninstaller.py  ──►  build_installer.py
   │                   │                         │
   ▼                   ▼                         ▼
pkg.zip          Uninstaller.exe        dist/Tlamatini_Release/

7.2. Step 1 — build.py

Installs deps, runs collectstatic, executes PyInstaller, copies required payloads (including README.md, the self-knowledge map Tlamatini.md, and bundled jd-cli/), runs migrations, creates the default user (user/changeme), renames the exe to Tlamatini.exe, copies all 76 agent templates, bundles support scripts (register_flw.ps1, CreateShortcut.ps1, Tlamatini.ps1, Tlamatini.ico), and zips it all into pkg.zip.

build.py is strict: missing README.md, missing jd-cli/, or missing jd-cli.bat causes a non-zero exit.

Self-modify builds. Add the --self-modify flag to ship Tlamatini's own source tree inside the distribution:

python build.py --self-modify

When the flag is present (self_modify = "--self-modify" in sys.argv), the build copies Tlamatini/agent/TlamatiniSourceCode/ recursively to the install root next to the exe, so it resolves like prompt.pmt / config.json / Tlamatini.md, and Tlamatini can read, inspect, and modify her own code at runtime. Without the flag the directory is omitted entirely. The build prints Self-modify build : YES (or no) so you can confirm which kind of build you produced. See §9.6 for how the LLM uses it.

7.3. Step 2 — build_uninstaller.py

python build_uninstaller.py

Builds uninstall.py into a single --onefile Tkinter exe. Output: Uninstaller.exe at the project root.

7.4. Step 3 — build_installer.py

python build_installer.py

Requires pkg.zip and Uninstaller.exe. Builds install.py with --onedir --windowed and a splash screen, copies pkg.zip and Uninstaller.exe into dist/Installer/, and assembles dist/Tlamatini_Release/ with SHA-256 verification.

The final distributable is dist/Tlamatini_Release/ — zip the folder, share it.

7.5. What the installer does on the end-user box

1. Tkinter GUI to choose installation directory (no admin needed).
2. Extracts pkg.zip into <install_path>/Tlamatini/.
3. Locks agent venv permissions.
4. Writes config.json.
5. Copies Uninstaller.exe.
6. Creates desktop and Start Menu shortcuts (Tlamatini.lnk — falls back to user-scoped paths under restrictive Group Policies).
7. Registers .flw to open with Tlamatini.
8. Cleans the PyInstaller bundle path from helper subprocess environments.

Frozen mode resolves config.json from the executable directory (or CONFIG_PATH env var). Template-agent discovery uses <install_dir>/agents in frozen mode and Tlamatini/agent/agents/ in source mode. _resolve_python_executable() tries PYTHON_HOME → bundled python.exe → PATH.

7.6. Source mode vs Frozen mode: why flows still work

Tlamatini has two operational modes:

The new Flow Compiler was built to respect both modes. It does not assume your repo is at C:/Development/Tlamatini, and it does not assume the installed app lives in a specific Program Files folder.

The compiler asks Tlamatini at runtime:

1. "Am I frozen?"
2. "Where are the agent templates?"
3. "Where is this user's session pool?"
4. "Which agent contract applies to this node?"

Then it writes only into the current pool:

agents/pools/<session_id>/<agent_name_n>/config.yaml

That path exists in both modes. In source mode it is under the repo's Tlamatini/agent/agents/pools/. In frozen mode it is under the installed app's agents/pools/.

For users, the takeaway is simpler: a .flw saved in source mode should load in an installed build, and a .flw saved in an installed build should load back in source mode.

8. Configuration (Tlamatini/agent/config.json)

8.1. LLM and unified-agent

{
  "embeding-model": "Nomic-Embed-Text:latest",
  "chained-model": "kimi-k2.6:cloud",
  "ollama_base_url": "http://127.0.0.1:11434",
  "ollama_token": "",
  "ANTHROPIC_API_KEY": "<ANTHROPIC_API_KEY goes here>",
  "GEMINI_API_KEY": "<GEMINI_API_KEY goes here>",
  "enable_unified_agent": true,
  "unified_agent_model": "kimi-k2.6:cloud",
  "unified_agent_base_url": "http://127.0.0.1:11434",
  "unified_agent_temperature": 0.0,
  "unified_agent_max_iterations": 4096,
  "chat_agent_limit_runs": 100
}

unified_agent_max_iterations caps the Multi-Turn tool loop (default 4096). enable_unified_agent is the master switch for tool-calling.

8.2. RAG

Key knobs: chunk_size (3000), chunk_overlap (800), k_vector / k_bm25 (100 each), k_fused (150), enable_bm25, rrf_k (60), max_doc_chars (150000), max_context_chars (250000), and a context_budget_allocation map (high_relevance: 0.60, architecture: 0.20, related: 0.15, documentation: 0.05). See BookOfTlamatini.md Part VII for the full schema.

8.3. ACPX

{
  "acpx": {
    "cwd": "C:/Development/Tlamatini",
    "stateDir": "C:/Users/<you>/.tlamatini/acpx-state",
    "probeAgent": "gemini",
    "permissionMode": "approve-reads",
    "nonInteractivePermissions": "deny",
    "timeoutSeconds": 180,
    "agents": {
      "claude": { "command": "C:/Users/<you>/AppData/Roaming/npm/claude.cmd",
                  "env": { "ANTHROPIC_API_KEY": "sk-ant-..." } }
    }
  }
}

permissionMode ∈ approve-reads (default) / approve-all (DANGEROUS) / deny-all. The whole acpx block is optional; on first boot of an upgrade build, boot_acpx() appends the documented default block atomically.

8.4. MCP services and other knobs

• mcp_system_server_port (8765), mcp_files_search_server_port (50051) — MCP daemons.
• internet_classifier_model, web_summarizer_model, web_context_max_chars — internet toggle.
• image_interpreter_model, image_interpreter_base_url — vision.
• history_summary_*, keep_last_turns — chat-history compression.
• kali_server_url (http://127.0.0.1:5000) — the MCP-Kali-Server address auto-injected into chat_agent_kalier (see §3.15).
• stm32_mcp_server_script (now "" — empty triggers zero-config auto-bootstrap), stm32_mcp_repo_url, stm32_mcp_install_dir — the STM32 Template Project MCP for STM32er (see §3.16). Leave stm32_mcp_server_script empty and STM32er downloads, installs, and validates the server itself on first use.
• pio_executable (now "" — empty triggers zero-config auto-bootstrap), pio_core_dir — PlatformIO Core for ESP32er (see §3.17). Leave pio_executable empty and ESP32er downloads, installs, and validates PlatformIO itself on first use.
• arduino_cli_executable (now "" — empty triggers zero-config auto-bootstrap), arduino_cli_install_dir — the Arduino CLI for Arduiner (see §3.18). Leave arduino_cli_executable empty and Arduiner downloads the arduino-cli binary, initializes it, and (when auto_core_install is on) installs the FQBN's core itself on first use.

You no longer need to hand-edit all of those values. On /agent/, open Config -> Models or Config -> URLs to edit the most common runtime knobs in-place. The browser validates model strings / URLs / hosts / ports, the backend validates again, and config_loader.save_config_updates() atomically merges only the changed keys into the active config.json. The same loader path is used in source mode and frozen builds, so the chat UI and the executable stop drifting onto different config copies.

9. Architecture at a Glance

9.1. Big picture

Browser (Chat / ACP Designer)
    │ WebSocket
    ▼
Django Channels (Daphne ASGI) → AgentConsumer
    │
    ├── RAG Pipeline (FAISS + BM25, RRF, context budgeting, OOM fallback)
    ├── Unified Agent (Multi-Turn loop, planner, wrapped runtimes)
    └── MCP Services (System-Metrics WS, Files-Search gRPC)
    │
    ▼
LLM Backends: Ollama | Claude API | Qwen vision     +     ACPX Runtime → external CLIs

9.2. The five layers

9.3. Multi-Turn execution pipeline

Frontend (toggles) → WebSocket → AgentConsumer → ask_rag() (skips prompt-shape validator)
  → UnifiedAgentChain.invoke() → filter_acpx_tools(tools, acpx_enabled)
    → planner picks ≤20 tools (capability scoring + history-aware boost)
      → MultiTurnToolAgentExecutor: 1..4096 iterations of (LLM call → tool calls → ToolMessage)
        → [if ask_execs_enabled] before each state-changing tool: BLOCK on a browser
          Proceed/Deny prompt (ExecPermissionBroker); Deny halts the chain
        → Exec Report HTML appended (if exec_report_enabled, BEFORE save_message)
          → [if denied] red "Execution interrupted" banner appended (always)
            → broadcast → frontend renders, shows Create Flow if all 4 gates pass

Ask Execs gate. When the Ask Execs toggle is on (Multi-Turn-only), the executor pauses before every state-changing Tool/MCP/Agent and blocks on a Proceed/Deny dialog in the browser. The bridge is agent/exec_permission.py::ExecPermissionBroker: the synchronous executor thread emits an exec_permission_request frame onto the consumer's event loop and waits on a threading.Event; the browser's exec-permission-response resolves it. A Deny halts the whole chain and surfaces a red banner naming exactly what was denied; the round-trip is fail-safe (emit failure / cancel / disconnect all resolve to deny).

9.4. Agent contracts and the Flow Compiler

Tlamatini has two ways to create flows:

1. The chat can infer a flow from Multi-Turn tool calls.
2. The ACP canvas can build a flow by dragging agents and drawing arrows.

Those two paths now meet at the same backend contract layer. This is important because flow files are not just pictures. A flow must eventually become a set of real agent folders, and every folder needs a correct config.yaml.

The contract layer is intentionally small:

For beginners, the rule is: the canvas or chat creates a flow idea, then the Flow Compiler turns that idea into executable agent folders.

The compiler does a few quiet but important safety jobs:

• It starts from each agent template's config.yaml, then merges only the node's custom settings.
• It clears and rebuilds managed connection fields, so stale arrows from an old pool do not survive by accident.
• It understands special agent wiring such as AND, OR, Asker, Forker, Counter, Ender, Stopper, and Cleaner.
• It writes interconnection-scheme.csv for Parametrizer nodes when mappings are saved in the .flw.
• It keeps FlowCreator and FlowHypervisor out of runtime validation because they are helper/control agents, not normal flow workers.
• It redacts known secrets for remote chat ingress agents such as TeleTlamatini and WhatsTlamatini when chat-created flows are exported.

This is the Pareto improvement: a small shared backend layer makes both major features safer. Chat-created flows and ACP-created flows now speak the same format before they touch the runtime.

9.5. Agent catalog (the 76 types, by family)

Per-agent details (config knobs, lifecycle, naming convention, log markers): see BookOfTlamatini.md Part IV — The Tlamatini Bestiary. To add a new agent, follow Tlamatini/.agents/workflows/create_new_agent.md (8-step checklist).

9.6. Self-Knowledge & Self-Modification

Tlamatini ships with a first-person self-knowledge map — Tlamatini/agent/Tlamatini.md — that the LLM reads as her own description of who and what she is: her two runtime modes (frozen vs source, and how to tell them apart), the ports she opens (8000 for the web app, 8765 for the System-Metrics MCP, 50051 for the Files-Search MCP), her main pages, her tech stack, her full capability surface, and how she can improve herself. The audience is the LLM alone, so the file deliberately does not follow prompt.pmt's HTML/contrast styling rules — it is a private self-reference, never rendered to users.

The map is injected into the system prompt at prompt-build time. prompt.pmt carries a <self_knowledge>{self_knowledge}</self_knowledge> block, and agent/rag/config.py fills it in: _load_self_knowledge_block() reads Tlamatini.md, brace-escapes it ({ → {{, } → }}) so its code snippets cannot collide with the f-string template variables, and fails open — a missing, empty, or unreadable file degrades to a short literal notice instead of raising. The substitution happens at the single prompt-load site in load_config_and_prompt(), so it covers all four chains (basic, history-aware, unified, prompt-only) without adding a new input variable. Tlamatini.md is resolved from the application directory exactly like prompt.pmt and config.json (the install root next to the .exe in frozen mode, Tlamatini/agent/ in source mode), and build.py ships it both via --add-data=…/Tlamatini.md;agent and by copying it to the install root so frozen resolution next to the exe works. The identity rules in prompt.pmt point the LLM at Tlamatini.md whenever a prompt concerns who or what she is, her architecture / modes / ports / pages / internals, or improving herself.

Self-modification is a second, independent capability axis. The optional directory Tlamatini/agent/TlamatiniSourceCode/, when present, contains Tlamatini's own source code so she can read, inspect, and modify herself — present means a self-able-modify build; absent means a not-self-able-modify build (orthogonal to frozen vs source). The tree is bundled only when build.py is invoked with the new --self-modify flag (see §7.2); without the flag it is omitted entirely. Because the directory is optional, prompt.pmt instructs the LLM to always verify the directory's presence (for example, a Multi-Turn directory listing) before claiming she can read or edit her own code; if it is absent she says so and falls back to the injected self-knowledge plus the docs.

When you load your own project as context (Context ▸ Set directory / Set file as context) and then ask a generic "summarize the project / the source code / the provided context" question, the loaded context takes priority over the always-injected self-knowledge — so Tlamatini summarizes your code, not herself. This is enforced by a prompt.pmt loaded-context-priority rule plus a deterministic scope header (agent/rag/utils.py::prepend_loaded_context_scope()) applied across all four chains.

10. Embedding-Memory Pre-Flight Guard (GPU hosts)

When you click Set directory as context in the Context menu, Tlamatini walks the directory, splits each file into chunks, and pushes every chunk through Ollama's embedding API to build a FAISS index. On a PC / consumer GPU a heavy embedding model can occupy 75–95% of total VRAM by itself — and once a chat model is also resident the combined footprint exceeds available memory and the daemon thrashes RAM↔VRAM swap on every embed batch. A 30-second context-load turns into a multi-hour stall.

The embedding-memory pre-flight guard (Tlamatini/agent/embedding_memory_guard.py) catches this before the embed burst starts. It runs only when an NVIDIA GPU is detected; on CPU-only / AMD / Apple Silicon hosts it is a silent no-op and the legacy load path is unchanged.

10.1. Why this exists

The trigger case is the dev box this codebase is calibrated on: an NVIDIA GeForce RTX 4070 Laptop GPU with 8 188 MiB of VRAM. The previously-configured embedding model qwen3-embedding:8b (7.6 B parameters, Q4_K_M quantization) sits at ~6.24 GB resident — 77.9% of total VRAM. Add the chat model and the daemon evicts something on every embed batch. The fix is either: switch to a smaller model (e.g. nomic-embed-text:v1.5 at ~0.60 GB resident, 7% saturation) or accept the slow path knowingly. The guard surfaces the choice before the heavy work starts, so you can abort, swap the model in config.json, and restart Ollama — saving an hour of debugging "why is context loading frozen?".

10.2. How the check fires (the hook point)

The guard is wired into agent/consumers.py::setup_contextual_rag_chain at exactly one point: after the "loading context" banner is broadcast to the chat, before the heavy asyncio.to_thread(setup_llm_with_context, …) call that drives the embedding burst. The flow is:

WebSocket "set-directory-as-context"
    ↓
consumers.py:setup_contextual_rag_chain(path_only)
    ↓
broadcast MSG_AGENT_LOADING_CONTEXT chat bubble
    ↓
► embedding_memory_guard.check_embedding_memory_for_directory(...)
    │
    ├─► returns None (no GPU / under threshold / probe failed)
    │       → proceed silently
    │
    └─► returns warning dict
            → broadcast HTML warning chat bubble
            → proceed anyway (informational, non-blocking)
    ↓
asyncio.to_thread(setup_llm_with_context, ...)
    └─► OllamaEmbeddings + FAISS.from_documents(...)   ← VRAM burst

The check runs inside asyncio.to_thread so a slow nvidia-smi or cold /api/show call never blocks the Channels event loop. The whole block is wrapped in try/except Exception so any unhandled probe error prints a one-line [EMBED-MEM] Pre-flight check skipped (fail-open) to tlamatini.log and the load continues — a diagnostic must never block the user.

10.3. GPU detection and no-GPU behavior

The guard reuses the already-cached nvidia-smi -L probe from agent/gpu_perf.py::_has_nvidia_gpu() (introduced for the model-pinning hook). The probe runs at most once per process; subsequent calls hit the in-module cache. On CPU-only Linux/Windows, AMD GPUs, and Apple Silicon the probe returns False once at server start and every subsequent call to the guard returns None immediately — no subprocesses spawned, no HTTP calls made, no overhead.

This is the portability guarantee: a fresh git pull on a no-GPU box behaves exactly as before the guard existed. The 28 dedicated no-GPU compatibility tests (see 10.7) lock the contract in place.

10.4. Three-tier VRAM prediction

When the GPU gate passes, the guard predicts the embedding model's resident VRAM in priority order:

Tier B uses a standard llama.cpp / GGUF bits-per-weight table:

Unknown quants fall back to a conservative 5.0 bits/weight. The overhead multiplier accounts for KV cache + activation buffers + GGML allocator slack:

• × 1.40 for models with ≥ 1 B parameters (large-model regime)
• × 2.20 for sub-1 B models (proportionally larger KV/buffer overhead)

Calibration against live measurements on the RTX 4070 Laptop:

Tier B also pulls the embedding dimension from /api/show via any architecture-prefixed *.embedding_length key (e.g. qwen3.embedding_length=4096, nomic-bert.embedding_length=768). Combined with a directory pre-scan that mirrors the exclusion rules of agent/rag/factory.py::CustomTextLoader, it reports a projected FAISS-index RAM size (num_chunks × embedding_dim × 4 bytes, float32). This is RAM, not VRAM, but useful to surface on directories with hundreds of thousands of chunks.

10.5. The 80% threshold and the chat-bubble warning

The guard fires when predicted_vram ≥ 0.80 × smallest-GPU total VRAM. Why the smallest GPU (rather than the sum or the largest)? Because Ollama loads each model into a single device — using the max would silently under-report the constraint on heterogeneous multi-GPU rigs.

When the threshold is crossed, the guard returns a structured dict the consumer renders as an HTML chat bubble. A real example from this dev box (artificially threshold-lowered to 70% so qwen3-embed:8b trips):

⚠️ Embedding-memory warning
Embedding model qwen3-embedding:8b needs ~6,378 MiB of VRAM
(currently resident in VRAM), which is 77.9% of the smallest
GPU's total (8,188 MiB) — above the safety threshold of 70%.
Projected FAISS vector store (RAM, not VRAM): ~28 MiB across
1,847 chunks at dim 4096.
Context loading will continue, but expect slow embedding
throughput or RAM↔VRAM swap. To eliminate the pressure, switch
embeding-model in config.json to a smaller model
(e.g. nomic-embed-text:v1.5) and restart.

The message is informational and non-blocking — context loading proceeds. The user picks whether to wait it out, hit Cancel, or change models. The phrasing names the exact config.json key (embeding-model, with the spelling preserved from the existing codebase) and a concrete alternative.

10.6. Tuning, overrides, and what the guard does NOT do

What the guard does NOT do, by deliberate choice:

• It does not abort context loading. The warning is informational. (If you want abort-on-warning behavior, wire a confirm/cancel WebSocket round-trip — the surface is described in agent_page_init.js near set-dir-context.)
• It does not estimate the chat model's VRAM. Only the embedding model is checked, because that is the model the directory-load path forces into VRAM. The chat model is handled by gpu_perf.pin_ollama_model separately.
• It does not persist warnings. Each context-load runs an independent check.
• It does not call nvidia-smi on CPU-only hosts. Both gates (_has_nvidia_gpu_cached and the _gpu_total_memory_bytes query) short-circuit before any subprocess spawn.
• It does not add new dependencies. subprocess, urllib.request, and os.walk are the only stdlib touchpoints — the same surface agent/gpu_perf.py already uses.

10.7. Test coverage

The guard ships with 49 automated tests in Tlamatini/agent/test_embedding_memory_guard.py, split into seven SimpleTestCase classes:

The NoGpuCompatibilityTests class is the portability proof. Its coverage matrix:

test_real_entry_point_call_never_raises is the CI gate: it makes the actual subprocess.run(["nvidia-smi", ...]) and urlopen("http://127.0.0.1:11434/...") calls against whatever the runner offers, and asserts the return is either None or a well-formed warning dict — never an exception. The same test passes on this RTX 4070 dev box (returns None because qwen3-embed sits at 77.9%, under the 80% gate) and on a CPU-only CI image (returns None because the GPU gate fails fast).

Run them yourself:

cd Tlamatini
python manage.py test agent.test_embedding_memory_guard --verbosity=2
# 49 tests in ~2.3 s, no DB setup, no GPU required.

11. Orphan-Process Cleanup (conhost.exe reaper)

Tlamatini now ships a three-tier reaper (Tlamatini/agent/orphan_reaper.py) that cleans up the console-host children every console subprocess on Windows drags behind it. Without this pass, users were occasionally seeing conhost.exe processes lingering in Task Manager with the Tlamatini icon — the icon is inherited from the parent EXE that spawned the console — and reasonably concluding that Tlamatini was leaking processes.

11.1. The problem this solves

On Windows, when a tool (execute_command, chat_agent_executer, an ACPX CLI child, an agent-pool Python subprocess, …) spawns a console child, Windows allocates a conhost.exe companion to host that console. If the immediate parent dies before the OS reaps the console pair, that conhost.exe outlives Tlamatini. Two compounding causes were closed at once:

• The reaper itself sweeps zombies and orphaned console hosts at three lifecycle points (below).
• Spawn sites were hardened — views.py::execute_starter_agent_view, execute_ender_agent_view, restart_agent_view, execute_flowcreator_view, every ACPX child in acpx/runtime.py, and a subprocess.Popen.__init__ guard at the top of agents/ender/ender.py (mirrored across every other pool agent) now spawn with CREATE_NO_WINDOW | DETACHED_PROCESS | CREATE_NEW_PROCESS_GROUP and stdio piped to DEVNULL. No console is allocated in the first place, so no conhost.exe companion exists to orphan.

11.2. The three tiers

11.3. What gets reaped (and what does not)

A process is considered a "Tlamatini orphan" if any of the following hold:

• It is a descendant of the current Tlamatini PID and its status is ZOMBIE / DEAD.
• It is a conhost.exe / openconsole.exe whose parent PID is in our process tree, OR whose parent PID no longer exists.
• Its cmdline references the agent-pool directory (agents/pools/... or agents/pools/_chat_runs_/...) but it is no longer tracked.

Each candidate is escalated terminate → wait 1 s → kill via psutil; an "unable-to-kill" outcome surfaces as a survivor, never as an exception. The reaper never raises into the caller — a cleanup that crashes the chat path would be worse than the orphans it tries to kill.

Out of scope on purpose: console hosts spawned by unrelated processes (a different IDE, your shell, another app's child) — the parentage check keeps the sweep narrow.

11.4. The user-visible follow-up message

When Tier 2 detects survivors, the user sees a second chat bubble immediately after the main answer:

⚠ Heads-up: Tlamatini tried to clean up after this request but the following
process(es) refused to terminate. They are most likely harmless leftovers from
a tool you ran, but if you do not recognize them please end them manually from
Task Manager so no Tlamatini-spawned child outlives the app:
  • conhost.exe — PID 18244
  • python.exe — PID 19108

The rendering helper is orphan_reaper.format_survivors_message(); it returns None (so no extra message is sent) when the survivor list is empty, which is the common case after the spawn-site hardening landed.

12. Troubleshooting

12.1. Ollama / models

• "connection refused" → ollama serve in a dedicated terminal. Check ollama_base_url.
• Model not found → ollama list to see what's pulled. Pull the missing tag.
• Remote Ollama → set ollama_token for bearer auth.

12.2. RAG / context

• Set-Context shows no green banner → check file permissions, ensure files are text not binary.
• "Out of memory" during embedding → fallback mode kicks in; retrieval quality drops, files still accessible. Switch to a smaller embedding model. See chapter 10 — the embedding-memory pre-flight guard now warns you about this before the embed burst starts on GPU hosts.
• Hit max_doc_chars → bump it.
• Session says it was restored after a refresh, but the input stays disabled briefly → that is expected while the contextual RAG chain rebuilds. Wait for the ready state / spinner to clear before sending the next prompt.

12.3. Multi-Turn / planner

• Did you tick Multi-Turn? Is enable_unified_agent: true?
• "Tool X is not available" → the planner did not bind X. Check [Planner._select] console lines, add matching keywords to your prompt, or raise max_selected_tools.
• 4096 iterations exhausted → likely a busy-poll loop. Use chat_agent_sleeper / chat_agent_run_wait instead.

12.4. Chat-created flows and ACP validation

• Create Flow downloads a .flw, but it looks simpler than the chat transcript. That is normal. The file stores the flow structure, node config, connections, and artifacts. It does not store the entire conversation.
• Create Flow fails from the chat. The browser first asks /agent/flow_from_tool_calls/ to normalize the draft. If that endpoint fails, the frontend falls back to the older browser-only export so you do not lose the flow draft.
• A TeleTlamatini or WhatsTlamatini flow is missing passwords or tokens after export. That is intentional. Known secret fields are redacted when the chat exports a flow. Re-enter secrets in the agent config before running the flow.
• Validate shows stale connections. Validate now asks /agent/compile_flow/ for a dry-run compile of the live canvas instead of trusting whatever is already in the pool. If the canvas still looks wrong, save the .flw, reload it, and check the browser console for a compile error.
• Start runs an older version of the flow. Start now compiles the visible canvas in write mode before launching. If you still see old behavior, clear the pool from the ACP close/clear controls, load the .flw again, and run Validate once before Start.
• A value I set in an agent dialog disappears after Validate or Start. Current builds preserve explicit dialog edits and merge canvas-derived wiring on top of them. If something still looks wrong, reopen the dialog, save once more, then run Validate and inspect the compile warnings.
• Parametrizer mappings disappear after reload. Save the flow after creating mappings. New .flw files store mappings under artifacts.parametrizerMappings, and the loader restores them into each Parametrizer node.

12.5. ACPX / external CLIs

• acp_doctor says agent not resolvable → CLI not on PATH, or set acpx.agents.<id>.command to the absolute path.
• Transcript only shows outbound prompts on Windows → your build is older than May 2026. Update — fix is transport="oneshot-prompt" for claude/gemini/cursor/qwen/codex.
• API key not picked up → per-agent acpx.agents.<id>.env wins over exported shell vars; check both.
• Session left running → always end with acp_kill. If a request times out, manually acp_list_sessions + acp_kill.

12.6. Frozen build / installer

• Wrong config used → place config.json next to the exe, or set CONFIG_PATH.
• Missing templates → verify agents/ exists in the install. Rebuild if README.md, jd-cli/, or template directories are missing.
• Restrictive Group Policy blocks shortcuts → CreateShortcut.ps1 falls back to user-scoped Desktop / Start Menu paths.

12.7. Logs to consult first

tlamatini.log silences successful GET / 200/304 polling lines but keeps non-2xx/3xx GETs visible.

INFO loggers worth knowing: agent.chat_agent_runtime, agent.tools, agent.mcp_agent, agent.global_execution_planner, agent.capability_registry.

13. Versioning

Tlamatini follows Semantic Versioning 2.0.0 — MAJOR.MINOR.PATCH — with git tags as the single source of truth. You never hand-edit a version string anywhere in the codebase; you tag, then you build, and the three build scripts bake the resolved value into every artefact.

13.1. The bump rules

Pre-releases use the standard SemVer suffixes — 2.0.0-alpha.1, 2.0.0-beta.1, 2.0.0-rc.1 — and sort before the final release.

13.2. Cutting a release

git tag -a v1.19.3 -m "Release 1.19.3: <one-line summary>"
git push origin v1.19.3
python build.py
python build_uninstaller.py
python build_installer.py

All three build scripts pick the tag up from git describe --tags automatically. The artefact lands in dist/Tlamatini_Release_v1.19.3/.

13.3. Where you can see the running version

All four are computed from the same Tlamatini/agent/_version.py that build.py writes (gitignored, regenerated on every build).

13.4. Building without tagging (development)

The build never fails for "no version" — and the version surface is always a clean SemVer like 1.1.1. The resolver returns the bare base tag reachable from HEAD; distance / commit / dirty state are deliberately stripped:

No .devN, no +gSHA, no .dirty ever appears in the version string — those concerns stay in git (git status, git describe --long --dirty).

13.5. Overriding the resolved version

build.py exports $env:TLAMATINI_VERSION after resolving, so build_installer.py and build_uninstaller.py in the same shell see the same value — the three artefacts cannot disagree.

The full contract — release recovery, runtime resolver internals, file-by-file integration map, FAQ — lives in VERSIONING.md.

14. Contributing & License

14.1. Contributing

1. Fork; create a feature branch.
2. Follow PEP 8. Run python -m ruff check and npm run lint before pushing.
3. Add tests in Tlamatini/agent/tests.py (the suite has 266+ tests; the bar is "zero regressions, ruff clean").
4. Update docs when API or behavior changes — this README, docs/claude/*.md, and agent/agents/flowcreator/agentic_skill.md for new agents.
5. Open a PR with a clear description.

When adding a new agent, follow the 8-step checklist in Tlamatini/.agents/workflows/create_new_agent.md (backend script + view + migration + CSS gradient + 4 JS files + agentic_skill.md + README + lint).

14.2. Acknowledgments

Django · LangChain · LangGraph · Ollama · FAISS · Anthropic · Bootstrap · TextMeBot · Ruff · PyAutoGUI · JD-CLI

14.3. License

GNU General Public License v3.0 — see LICENSE.

For the long-form, narrative documentation (full agent bestiary, complete WebSocket / HTTP API reference, glossary, full changelog, architecture deep dives), see BookOfTlamatini.md. For support, open an issue on GitHub.