How ganesh-dark/bioweaver works
Bioweaver accepts biomedical questions through a chat interface or API, then a ReAct agent selects database tools or answers conceptual questions with an LLM. Tool results are retrieved from external biological services or local datasets, optionally analyzed and combined, and returned as a bounded, paginated answer. The README documents the major domain capabilities; source excerpts and the SOURCE INDEX confirm key imports and calls, but not every tool’s internal wiring.
Main entry points: agent_graph.py ReAct Agent · reactome_client.py Reactome Queries · app.py Chat API · kegg_client.py KEGG Queries
Components 24
User Interface
- Chat Frontend
static/index.html - Chat API
app.py
Agent and Models
- ReAct Agent
core/agent_graph.py - Model Provider
core/ai_model_wrapper.py - ClinVar Query Router
clients/hf_router_client.py
Pathway and Metabolism
- KEGG Queries
clients/kegg_client.py - Reactome Queries
clients/reactome_client.py - KEGG Analytics
analytics/kegg_analytics.py - Reactome Analytics
analytics/reactome_analytics.py - Reaction Utilities
clients/reactome_reactions.py
Gene and Clinical Data
- ClinVar Search
clients/clinvar_client.py - NCBI Gene Lookup
clients/ncbi_client.py - HGNC Name Cache
clients/hgnc_client.py - Disease Overviews
clients/wiki_client.py
Data Sources
- KEGG TSV Data
Kegg_files - Reactome Data
Reactome_files - HGNC Dataset
hgnc_files/hgnc_complete_set.txt - ClinVar Database
Other components
- Researcher
- KEGG Service
- Reactome Service
- NCBI E-utilities
- Wikipedia
- LLM Provider
Connections 26
- Researcher → Chat Frontend: asks question
- Chat Frontend → Chat API: submits question
- Chat API → Chat Frontend: streams answer
app.py - Chat API → ReAct Agent: streams query
app.py - ReAct Agent → Model Provider: reasons with
core/agent_graph.py - Model Provider → LLM Provider: requests completion
- ReAct Agent → ClinVar Query Router: routes variants
core/agent_graph.py - ClinVar Query Router → Model Provider: extracts query fields
clients/hf_router_client.py - ReAct Agent → KEGG Queries: queries KEGG
core/agent_graph.py - ReAct Agent → Reactome Queries: queries Reactome
core/agent_graph.py - ReAct Agent → KEGG Analytics: runs analysis
core/agent_graph.py - ReAct Agent → Reactome Analytics: runs analysis
core/agent_graph.py - ReAct Agent → Reaction Utilities: retrieves reactions
- ReAct Agent → ClinVar Search: searches variants
core/agent_graph.py - ReAct Agent → NCBI Gene Lookup: looks up genes
core/agent_graph.py - ReAct Agent → Disease Overviews: gets overviews
core/agent_graph.py - KEGG Queries → KEGG Service: requests records
clients/kegg_client.py - Reactome Queries → Reactome Service: fetches participants
clients/reactome_client.py - NCBI Gene Lookup → NCBI E-utilities: searches genes
clients/ncbi_client.py - KEGG Analytics → KEGG TSV Data: reads TSVs
analytics/kegg_analytics.py - Reactome Queries → Reactome Data: loads pathway model
- ClinVar Search → ClinVar Database: queries variants
clients/clinvar_client.py - KEGG Queries → HGNC Name Cache: uses gene names
clients/kegg_client.py - Reactome Queries → HGNC Name Cache: expands symbols
clients/reactome_client.py - HGNC Name Cache → HGNC Dataset: loads gene names
clients/hgnc_client.py - ReAct Agent → Wikipedia: falls back for overview
README.md