How pranav1237/qyravex works
QYRAVEX is an auditable financial-research agent, not a trading system. Users submit a research question through the CLI or browser; a bounded planner selects typed research tools, which retrieve and calculate evidence. The agent records evidence and run metrics, then the CLI or dashboard presents a synthesized report with provenance, conflicts, and limitations. The README also documents a browser dashboard, market charting, forecasting, and offline benchmarks. Sampled source supports the core orchestration; some dashboard and forecast internals were not inspected.
Main entry points: tools.py Tool Registry · web.py Web API · agent.py Research Agent · cli.py Research CLI
Components 19
User Interfaces
- Research CLI
src/workbridge/cli.py - Web API
src/workbridge/web.py - Browser Dashboard
src/workbridge/static/app.js
Research Workflow
- Research Agent
src/workbridge/agent.py - Tool Registry
src/workbridge/tools.py - Report Synthesis
src/workbridge/synthesis.py - Research Models
src/workbridge/models.py
Data Sources
- Provider Clients
src/workbridge/providers.py - SEC EDGAR
- Market Data
- FRED Data
- Public News
- Finnhub
- NewsAPI
Evidence and State
- Research Memory
src/workbridge/memory.py - Runtime Settings
src/workbridge/config.py
Quality and Monitoring
- Run Metrics
src/workbridge/observability.py - Offline Benchmarks
src/workbridge/benchmarks.py
Other components
- Researcher
Connections 31
- Researcher → Research CLI: submits research
README.md - Researcher → Browser Dashboard: submits research
README.md - Browser Dashboard → Web API: uses API
README.md - Research CLI → Research Agent: runs research
src/workbridge/cli.py - Research CLI → Tool Registry: creates registry
src/workbridge/cli.py - Research CLI → Research Memory: opens memory
src/workbridge/cli.py - Research CLI → Report Synthesis: synthesizes report
src/workbridge/cli.py - Web API → Research Agent: runs research
src/workbridge/web.py - Web API → Tool Registry: creates registry
src/workbridge/web.py - Web API → Research Memory: uses memory
src/workbridge/web.py - Web API → Report Synthesis: synthesizes report
src/workbridge/web.py - Web API → Offline Benchmarks: runs challenges
src/workbridge/web.py - Web API → Provider Clients: gets chart data
src/workbridge/web.py - Research Agent → Tool Registry: executes tools
src/workbridge/agent.py - Research Agent → Research Memory: stores evidence
src/workbridge/agent.py - Research Agent → Run Metrics: records metrics
src/workbridge/agent.py - Research Agent → Research Models: uses models
src/workbridge/agent.py - Tool Registry → Provider Clients: retrieves data
src/workbridge/tools.py - Tool Registry → Research Memory: searches evidence
src/workbridge/tools.py - Tool Registry → Research Models: returns results
src/workbridge/tools.py - Provider Clients → Runtime Settings: reads settings
src/workbridge/providers.py - Tool Registry → SEC EDGAR: retrieves filings and facts
README.md - Tool Registry → Market Data: retrieves prices
README.md - Tool Registry → FRED Data: retrieves macro series
README.md - Tool Registry → Public News: searches headlines
README.md - Provider Clients → Finnhub: optional enrichment
README.md - Provider Clients → NewsAPI: optional enrichment
README.md - Report Synthesis → Research Models: builds report
src/workbridge/synthesis.py - Offline Benchmarks → Research Memory: checks persistence
src/workbridge/benchmarks.py - Offline Benchmarks → Report Synthesis: checks synthesis
src/workbridge/benchmarks.py - Offline Benchmarks → Research Models: checks evidence models
src/workbridge/benchmarks.py