IAF v0.9 — First Public Release

The Fundamental Loop is live. Dispatch and Scheduler are next.

2026-03-30 v0.9

What is IAF?

Intelligenism Agent Framework is a ~550-line Python agent framework where each agent owns a complete, independent copy of the engine. Built on Intelligenism theory, it practises possibility management over certainty management — no single collaboration paradigm is hardcoded into the framework. Adding an agent is copying a folder. Adding a collaboration mode is adding a folder. Complexity grows additively, never multiplicatively.

What ships in v0.9

Agent Layer — Complete

The core of IAF: the Fundamental Loop. Each agent is a self-contained runtime with its own LLM engine, tool executor, context manager, identity file (SOUL.md), and skill system. Current implementation: ~461 lines for the core loop + 89 lines of shared infrastructure.

LLM Communication — API connectivity via OpenRouter (Gemini Flash and others), message assembly, structured response handling
Tool Executor — Auto-discovery registry, whitelist-based execution, tool call loop
Context Management — Sliding-window trimming strategy, chat history persistence
Identity & Skills — SOUL.md persona definition, trigger-based skill injection

UI Layer — Complete

A browser-based interface following the yellow-pages architecture:

index.html — Auto-discovers all available pages and agents
chat.html — Agent selector + conversation window
pages/ — Drop an HTML file, refresh, it appears. Zero coupling between pages.

Shared Infrastructure

lib/llm_client.py — HTTP calls + retry logic + error classification (76 lines)
lib/token_utils.py — Token estimation (13 lines)
config.json — Multi-provider configuration (OpenRouter, Anthropic, xAI)

What is in progress

Dispatch Layer — Architecture Designed

Multi-agent collaboration orchestration. Each collaboration strategy will be an independent folder under dispatch/, containing its own orchestration logic, context injector, config, and optional UI page. Dispatch calls lib/llm_client.call_llm() directly — zero coupling with agent engines.

Key design decisions already made:

Agents are data sources, not executors — Dispatch reads SOUL.md, skills, and model config from agent folders, but never calls agent engines
Three-layer context isolation — Agent chat history and Dispatch session context never pollute each other
Per-strategy configuration — Same agent can use different models in different collaboration strategies

Scheduler Layer — Architecture Designed

A unified cron-based trigger that scans schedules.json in both agent and dispatch folders. Fires single-agent tasks or multi-agent collaborations on schedule. Independent of chat_server.py — can run standalone for unattended operation.

Status Summary

✅ Fundamental Loop (Agent Layer) — complete
✅ UI Layer (Yellow Pages + Chat) — complete
✅ Shared Infrastructure (llm_client + token_utils) — complete
✅ Architecture Validation Tests — complete
🔧 Dispatch Layer — architecture designed, building now
🔧 Scheduler Layer — architecture designed, building now

Tech Stack

Pure Python + HTML. No Node.js, no npm, no frontend build tools.

pip install flask requests — that is the entire dependency list.

Get the Code

git clone https://github.com/IntelligenismCommercialDevelopment-LLC/intelligenism-agent-framework.git
cd intelligenism-agent-framework
pip install flask requests
# Configure your LLM provider in config.json
python chat_server.py
# Open http://localhost:5000