belief-engine 2.3.0

Autonomous AI build system. Describe what you want. Belief builds it, tests it, deploys it, and learns.

🧠 Belief Engine

An autonomous AI system that turns a sentence into working, tested software.

Describe what you want. Belief Engine builds it, tests it, deploys it, and learns from every build.

pip install belief-engine

belief --goal "Build a bookmark manager API with FastAPI — CRUD with tags, GET /random. SQLite." \
  --deploy docker_local

---

Benchmark: 85% Pass Rate

Tested on 20 challenges spanning single-file scripts to workflow DAG engines.

Pass rate:     17/20 (85%)
Avg weighted:  0.86
Cost per build: $0.18
Build time:    ~5 minutes

Tier 1 (scripts):        2/3
Tier 2 (CLIs + APIs):    4/4  ← perfect
Tier 3 (CRUD apps):      4/5
Tier 4 (multi-component): 3/4
Tier 5 (complex systems): 4/4  ← perfect

The engine builds complex systems (workflow engines, inventory managers, quiz platforms) more reliably than simple scripts. Tier 5 has been at 100% for three consecutive benchmark runs.

How It Works

You: "Build a todo app with Click"
  ↓
11 AI agents collaborate in a convergence loop:
  intake → research → planner → architect → skeleton → builder
  → covenant enforce → import fix → tester → executor → debugger
  → synthesizer → validator (real pytest) → water cycle → deploy
  ↓
Working software, tested, Dockerized, deployed.

The engine doesn't just generate code — it builds, tests, debugs, deploys, and learns. Every build deposits knowledge into ChromaDB soil. Patterns, antipatterns, and covenants feed future builds. Build 50 is smarter than build 1.

Key Numbers

Metric	Value
Codebase	76 Python files, ~19,800 lines
Benchmark	17/20 (85%) on 20-challenge suite
Builds completed	53+
Nutrients learned	140+
Self-learned covenants	7
Cost per build	$0.18 (was $0.87 — 80% reduction)
Build time	~5 minutes
LLM calls in validator	0 (fully deterministic)

What Makes This Different

It Learns From Every Build

ChromaDB-backed metabolization. Patterns, antipatterns, skeletons, and covenants accumulate in "soil" with FSRS confidence decay. After failed builds, the engine generates verbal self-critiques (Reflexion) that inform future similar builds.

Incompleteness Drives Convergence

Latios finds what's missing. Latias protects what matters. The tension between them drives builds forward — the "remainder" after each operation seeds the next.

Covenants Are Structural, Not Suggestions

Self-learned rules enforced via AST validators — not prompt injection. When the engine learned that from __future__ import annotations breaks SQLAlchemy's Mapped types, it added a deterministic AST check that removes the offending line automatically. Zero LLM tokens. Permanent fix.

Real Tests, Not Imagination

The validator runs actual pytest in a sandbox. Real pass/fail. Weighted scoring: smoke tests = 3x weight, functional = 2x, edge cases = 1x, environment errors = 0x.

SEED Self-Improvement

Every 5 builds, the engine analyzes its own failure patterns and proposes improvements. HIGH confidence proposals targeting prompt and configuration files are auto-applied with rollback on failure.

Quick Start

pip install belief-engine

# Set your API key
export ANTHROPIC_API_KEY=sk-ant-...

# Build something
belief --goal "Build a URL shortener with FastAPI and SQLite"

# Build + deploy
belief --goal "Build a REST API" --deploy docker_local --deploy-name myapi

# Run the benchmark
python3 -m belief.benchmark --tier=1,2,3,4,5

From Source

git clone https://github.com/metafiopy-tech/belief-engine.git
cd belief-engine
pip install -e ".[dev]"

Architecture

belief/
  agents/          — 11+ LangGraph agents (intake → validator)
  validators/      — AST covenant enforcers (deterministic, zero LLM)
  memory/          — ChromaDB metabolization (nutrients, soil, reflexion)
  refinement/      — Water cycle (analyze → fix → revalidate)
  deploy/          — Docker + Railway deployment
  codebase/        — Brownfield support (localization, patcher)
  languages/       — Multi-language adapters (Python, TypeScript)
  evolution/       — SEED self-improvement engine
  polarity/        — Latios/Latias incompleteness engine
  models/          — Pydantic models (state, artifacts, skeleton, contracts)
  hardening.py     — Budget limits, rate limiter, security scanner, audit log
  graph.py         — LangGraph pipeline wiring
  llm.py           — Anthropic API client with prompt caching + JSON repair

Model Routing

Agent	Model	Role
Research, Planner, Architect, Builder, Debugger	Sonnet 4.6	Deep reasoning
Intake, Tester, Gap Analyst, Synthesizer, Validator, Latios	Haiku 4.5	Mechanical tasks
Skeleton, Covenant Enforcer, Import Fix, Validator core	None	Deterministic (zero tokens)

Prompt caching provides 90% savings on repeated system prompts. Combined with Haiku routing, builds cost $0.15-0.25.

The Benchmark Suite

20 challenges across 5 tiers of complexity:

Tier	Challenges	Examples
1	3	FizzBuzz, Fibonacci, word count
2	4	Todo CLI, health API, calculator CLI, CSV stats
3	5	URL shortener, bookmark API, notes API, expense tracker, contacts
4	4	Blog engine, Kanban board, file vault, poll system
5	4	Event system, inventory manager, quiz engine, workflow DAG

Tech Stack

• Python 3.11+ (tested on 3.14)
• LangGraph for agent orchestration
• Anthropic Claude (Sonnet 4.6 + Haiku 4.5)
• ChromaDB for learning memory
• Docker for deployment

License

MIT

Author

Built by Fio — solo, from scratch, while making pizzas.

"The remainder after every operation drives the next cycle."