🏆 BioSwarm Engine v3.5

Production-ready multi-source intelligence swarm for AI-powered research

Overview • Features • Installation • Usage • Configuration • API • Roadmap

---

📸 Preview

$ bioswarm run --query "latest AI technology trends 2026" --depth 3

🏆 BioSwarm v3.5 - Starting intelligence swarm...
✅ SQLite database initialized
✅ Search clients ready
🤖 Spawning specialist agents...
✅ Swarm execution complete!
📊 Execution summary ready
📁 Exports written: Markdown, JSON, HTML, CSV

---

📋 Overview

BioSwarm Engine is a production-ready, multi-source intelligence CLI built in Rust. It orchestrates specialist AI research flows to gather fresh intelligence, persist execution history, and export usable reports.

Current strengths:

• ⚡ Production CLI flows for run, resume, export, history, and status
• 🔄 Resumable runs with SQLite-backed checkpoints
• 📊 Multi-format exports in Markdown, JSON, HTML, and CSV
• 🎯 Confidence scoring and executive summary generation
• 🗄️ Persistent history with searchable run records and trend deltas
• 🧪 Tested baseline with integration and snapshot coverage

---

✨ Features

Feature	Description	Status
🤖 Swarm Research Execution	Multi-agent business intelligence orchestration	✅
🗄️ SQLite Persistence	Schema migrations, checkpoint/resume, run history	✅
📤 Multi-Format Exports	Markdown, JSON, HTML, CSV with report templates	✅
🔄 Resume Flow	Recover and continue from stored checkpoints	✅
📈 Trend Analysis	Delta summaries across multiple runs	✅
🎯 Confidence Scoring	Per-run and aggregated confidence signals	✅
🧹 Deduplication	Duplicate insight cleanup in output generation	✅
🔍 Fresh Research Support	Fireworks generation with optional Exa freshness layer	✅
🛡️ Schema Migrations	Automatic database upgrades on startup	✅
🧪 Testing	Integration tests + snapshot tests	✅
🌐 REST API Server	HTTP entrypoint for automation	✅
🚀 Provider-flexible model backends	Fireworks, Ollama, OpenAI-compatible runtime switching	✅

---

🚀 Installation

Build from Source

git clone https://github.com/sayuru-akash/bioswarm-engine.git
cd bioswarm-engine
cargo build --release
cargo test

Prerequisites

• Rust 1.75+ (Install)
• API keys:
  • Fireworks AI for text generation
  • Exa AI optional, for fresher research/search enrichment

---

🎮 Usage

Basic Run

export FIREWORKS_API_KEY="your_key_here"
bioswarm run --query "AI market trends 2026"

Advanced Usage

bioswarm run \
  --query "sri lanka fintech opportunities" \
  --depth 3 \
  --formats markdown,json,html,csv \
  --output-dir ./reports \
  --database-path ./bioswarm.db

bioswarm history --limit 20
bioswarm status
bioswarm export --execution-id "<id>" --formats markdown,json
bioswarm resume --execution-id "<id>"

Backend-flexible runtime examples

# Fireworks
export FIREWORKS_API_KEY="fw_..."
bioswarm run --query "AI market trends" --backend fireworks --model accounts/fireworks/models/kimi-k2-instruct

# Ollama
export OLLAMA_API_KEY="local-or-placeholder"
bioswarm run --query "AI market trends" \
  --backend ollama \
  --model kimi-k2.5:cloud \
  --api-base-url http://127.0.0.1:11434/v1 \
  --api-key-env OLLAMA_API_KEY

# OpenAI-compatible
export OPENAI_API_KEY="sk-..."
bioswarm run --query "AI market trends" \
  --backend openai-compatible \
  --model gpt-4.1-mini \
  --api-base-url https://api.openai.com/v1 \
  --api-key-env OPENAI_API_KEY

CLI Commands

Command	Description
run	Execute a new swarm
resume	Resume from a saved checkpoint
export	Export an existing run
history	List past runs
status	Show current status and checkpoints

---

⚙️ Configuration

Environment Variables

# Required
export FIREWORKS_API_KEY="fw_..."

# Optional
export EXA_API_KEY="exa_..."
export DATABASE_PATH="./bioswarm.db"
export OUTPUT_DIR="./outputs"
export RATE_LIMIT_RPM=60
export BIOSWARM_DEPTH=2
export RUST_LOG=info

Config File (bioswarm.toml)

fireworks_api_key = "fw_..."
exa_api_key = "exa_..."
rate_limit_rpm = 60
database_path = "bioswarm.db"
output_dir = "outputs"
depth = 2
formats = ["markdown", "json", "html", "csv"]
backend = "fireworks"
model = "accounts/fireworks/models/kimi-k2-instruct"
api_base_url = "https://api.fireworks.ai/inference/v1"
api_key_env = "FIREWORKS_API_KEY"

---

🌐 API

BioSwarm includes an HTTP API server for automation:

./target/release/bioswarm-server

Endpoints

Endpoint	Method	Description
/health	GET	Health check
/api/v1/swarm/run	POST	Execute swarm
/api/v1/swarm/status	GET	Get current status/checkpoints

---

🏗️ Architecture

CLI / API
   ↓
Orchestrator
   ↓
Agent execution + checkpointing
   ↓
Search + generation clients
   ↓
SQLite persistence + export pipeline

Core files:

• src/main.rs CLI entry
• src/server.rs API server
• src/database.rs persistence + migrations
• src/orchestrator.rs swarm execution + checkpointing
• src/exports.rs export writers
• tests/ integration and snapshot coverage

---

🛣️ Roadmap

Planned next improvements:

• richer provider-specific capability negotiation
• stronger retry/backoff behavior for live provider calls
• richer report presentation and executive dashboards
• stronger CI/release polish
• deeper structured extraction from model outputs

Important: roadmap items above are next-step improvements beyond the currently shipped backend-flexible runtime support.

---

📜 License

MIT