ConsensusMind

Autonomous AI researcher for blockchain consensus mechanisms.

Overview

ConsensusMind is a local-first autonomous research toolchain for blockchain consensus protocols. It performs literature review, generates hypotheses, runs simulations/experiments, and produces publication-ready outputs (LaTeX paper + Markdown whitepaper).

Status

Current Version: 0.10.0

Project Status: Complete (Milestones 1–10 delivered)

Completed Milestones

Milestone 1: Foundation & Infrastructure

• Project initialization and structure
• Configuration system with TOML and environment variable support
• Logging infrastructure (file + console)
• LLM client with exponential backoff retry logic
• Integration tests
• Production-ready code quality (zero warnings)

Milestone 2: Knowledge Ingestion

• arXiv API integration with HTTPS
• PDF download and local storage
• JSON-based metadata tracking
• PDF text extraction and analysis
• Rate limiting and error handling
• Search and store workflow

Milestone 3: Knowledge Base

• Local embedding index build/update
• Semantic search over indexed papers

Milestone 4: Agent Core

• Agent runner with step tracking and persisted run memory
• Report generation

Milestone 5: Consensus Simulator

• Deterministic simulators with CLI entrypoints

Milestone 6: Hypothesis Generation

• LLM-driven hypothesis generation with novelty scoring
• Hypotheses persistence and reporting

Milestone 7: Automated Experimentation

• Experiment runner tied to hypothesis IDs
• Results saved under data/experiments and summarized into output reports

Milestone 8: Paper Generation

• LaTeX paper generation from hypotheses and experiment results

Milestone 9: Integration & Polish

• Unified CLI entrypoints (help/version and command usage)
• Repo structure improvements for outputs and data directories

Milestone 10: Whitepaper & Research Paper

• Whitepaper (Markdown) and research paper (LaTeX) generation from hypotheses and experiment results

Features

Current (v0.10.0)

• Configuration management from TOML files
• Environment variable overrides for sensitive data
• Structured logging to file and console
• HTTP client for vLLM/RunPod inference endpoints
• Automatic retry with exponential backoff
• Comprehensive error handling
• arXiv paper search and retrieval
• PDF download with duplicate detection
• Metadata persistence in JSON
• PDF text extraction for analysis
• Local embedding index + semantic search
• Agent run pipeline (search, download, index, retrieve, summarize, report)
• Hypothesis generation and persistence
• Consensus simulation and experimentation
• LaTeX paper generation from experiment outputs
• CLI help/version and stable command interface
• Whitepaper + research paper publishing commands

Architecture

Built in Rust for production reliability and performance.

Core Components:

• Agent executor with planning and memory
• Knowledge base with vector search
• Consensus protocol simulator
• LLM client for reasoning tasks
• LaTeX/Markdown output generation
• arXiv integration for paper retrieval
• PDF parsing and text extraction

Tech Stack:

• Language: Rust 2021 edition
• Async Runtime: Tokio
• HTTP Client: Reqwest with rustls
• Logging: Tracing
• Config: TOML
• LLM: Self-hosted vLLM (DeepSeek/Qwen)
• PDF Processing: pdf-extract
• Data Storage: JSON metadata + local files

Requirements

• Rust 1.70+
• GPU inference server (RunPod, self-hosted vLLM, or compatible endpoint)
• Storage for paper corpus

Installation

git clone https://github.com/ChronoCoders/consensusmind.git
cd consensusmind
cargo build --release

Configuration

Create config.toml in the project root with LLM endpoint, model settings, paths, agent parameters, and logging configuration.

For secrets (like API keys), use either:

• Environment variables (recommended), or
• config.local.toml (kept out of git via .gitignore)

Environment variable overrides available:

• LLM_ENDPOINT
• LLM_API_KEY
• LLM_MODEL
• CONFIG_PATH

Additional settings:

• knowledge.max_pdf_bytes controls the maximum allowed PDF download size.

Usage

consensusmind run "<query>"
consensusmind hypothesize "<query>"
consensusmind experiment <hypothesis-id> [--seeds N] [--ticks T] [--nodes N]
consensusmind paper <hypothesis-id>
consensusmind whitepaper <hypothesis-id>
consensusmind publish <hypothesis-id>
consensusmind index
consensusmind semantic-search "<query>" [top_k]
consensusmind simulate [rounds] [leader_failure_prob] [seed]
consensusmind raft-simulate [nodes] [ticks] [seed]
consensusmind help
consensusmind --version

Supports end-to-end research runs, hypothesis generation, experiments, and paper/whitepaper generation.

Outputs

Artifacts are written to disk so runs are reproducible and auditable:

• data/metadata.json: paper metadata store
• data/embeddings/index.json: local embedding index
• data/hypotheses.json: generated hypotheses
• data/experiments/<hypothesis-id>/results.json: experiment results
• output/reports/*.json: run/hypothesis/experiment reports
• output/papers/*.tex: generated LaTeX papers
• output/papers/*.md: generated Markdown whitepapers

Development

cargo build          # Build debug
cargo test           # Run tests
cargo fmt            # Format code
cargo clippy         # Lint
cargo build --release # Build optimized

This repository does not ship GitHub Actions workflows. Run the commands above locally for formatting, linting, and tests.

Roadmap

• Milestone 1: Foundation & Infrastructure
• Milestone 2: Knowledge Ingestion
• Milestone 3: Knowledge Base
• Milestone 4: Agent Core
• Milestone 5: Consensus Simulator
• Milestone 6: Hypothesis Generation
• Milestone 7: Automated Experimentation
• Milestone 8: Paper Generation
• Milestone 9: Integration & Polish
• Milestone 10: Whitepaper & Research Paper

License

Apache 2.0 - See LICENSE file

Contact

Distributed Systems Labs, LLC

• GitHub: https://github.com/ChronoCoders/consensusmind
• Website: https://dslabs.network

Contributing

This project maintains strict code quality standards:

• Zero compiler warnings
• Zero dead code
• Zero unused imports
• Production-ready quality required

Contributions welcome via pull requests.