Forge Modules

Each forge module extends the base pipeline with domain-specific scanners, belief templates, derivation rules, and review criteria.

Built-in modules

Module	Input	Best for
document	PDFs, markdown, code files	General-purpose belief extraction with section chunking
code	Git repos, source files	Architecture beliefs, technical decisions, API contracts
project	Issue trackers, PRs	Project state, decisions, blockers, team knowledge
product	GitHub/GitLab/Jira, docs	Product beliefs, user needs, feature priorities
meta	Other forge outputs	Cross-domain synthesis, contradiction detection, executive summaries

Document Forge

General-purpose belief extraction from any document format, with section-based chunking for PDFs and large documents.

reasonsforge document --sources-dir ./docs
reasonsforge document --pdf paper.pdf

Scanners: Reads markdown, text, HTML, and PDF files. PDFs are chunked by section headings (Abstract, Introduction, numbered sections, etc.) before summarization.

Best for: Converting documents into a queryable, justified knowledge base. Handles academic papers, technical documentation, and any text-heavy source.

Code Forge

Analyzes codebases to extract architectural beliefs, technical decisions, and API contracts.

reasonsforge code --repo ./my-project

# Step by step
reasonsforge code scan --repo ./my-project
reasonsforge code explore --loop 50
reasonsforge code propose-beliefs --auto
reasonsforge code derive --exhaust --auto

Scanners: Reads source files, parses function signatures, analyzes git history for decision patterns.

Best for: Understanding a codebase’s architecture, tracking technical decisions, onboarding new developers.

Subcommands: scan, explore, explain, walk-commits, propose-beliefs, accept-beliefs, review-proposals, verify, derive, topics, status, update

Project Forge

Extracts project state from issue trackers and pull requests.

reasonsforge project --github owner/repo

# Step by step
reasonsforge project init --github owner/repo
reasonsforge project scan
reasonsforge project explore --loop 20
reasonsforge project sprint-plan

Scanners: Pulls issues, PRs, comments, labels, and milestones from GitHub, GitLab, or Jira.

Best for: Project status dashboards, decision tracking, identifying blockers and dependencies, sprint planning.

Subcommands: init, scan, explore, propose-beliefs, accept-beliefs, review-proposals, research, derive, review-beliefs, repair, summary, sprint-plan, topics, status, update

Product Forge

Analyzes product data to extract user needs, feature priorities, and product beliefs.

reasonsforge product --github owner/repo --domain "Payment platform"

# Ingest product documents
reasonsforge product ingest docs/ --glob-pattern "**/*.md"

Scanners: Pulls issues with product labels, user feedback, feature requests, bug reports. Can also ingest markdown documents for product analysis.

Best for: Product roadmap analysis, user need identification, competitive analysis.

Subcommands: init, scan, ingest, explore, propose-beliefs, accept-beliefs, review-proposals, derive, generate-summary, summary, topics, status, update

Meta Forge

Cross-domain reasoning across expert belief networks. Unlike other forges, meta does not scan source material directly. It imports pre-built belief networks from other forges and finds emergent insights invisible to any single expert.

# Initialize with expert repos
reasonsforge meta init code=/path/to/code-expert project=/path/to/project-expert

# Full pipeline: import, derive, detect contradictions, summarize
reasonsforge meta update

# Ask questions across all domains
reasonsforge meta ask "What are the top cross-domain risks?"

How it works: Imports beliefs from other forges via sync_agent, namespacing them by source (e.g., code:belief-id, project:belief-id). Cross-domain derivations require antecedents from 2+ different agents. Contradiction detection finds beliefs from different domains that cannot all be true simultaneously.

Best for: Executive synthesis, cross-domain risk assessment, detecting contradictions between expert perspectives.

Subcommands: init, import, derive, ask, contradictions, summary, topics, status, update

Choosing a module

Pick the module that matches your data source:

• Have documents or PDFs? Use document
• Have a codebase? Use code
• Have an issue tracker for project management? Use project
• Have an issue tracker for product data? Use product
• Have multiple forge outputs to synthesize? Use meta

Each module shares the same underlying pipeline — the difference is in the scanners (what data it pulls) and the prompts (how beliefs are structured for that domain).