abhishsing/practo-support-agent

How abhishsing/practo-support-agent works

Practo is an offline clinic-support application for policy questions and appointment-status queries. Requests enter through HTTP or WebSocket endpoints, pass privacy, injection, budget, and caching controls, then use a sequential agent crew to retrieve policy evidence and/or look up appointment records; grounded drafts may receive a separate review before schema-validated responses are returned. The repository also supports policy-document ingestion and indexing. Relationships reflect the excerpts and README; unsampled implementation details are represented only where documented.

Main entry points: crew.py Sequential agent crew · main.py HTTP and chat API · index.py Policy indexer · logging_utils.py Masked event logging

Components 18

Request handling

HTTP and chat API api/main.py
Masked event logging api/logging_utils.py

Governance and review

Request guardrails agents/guardrails.py
Query budget governance/policy.py
Generation cache governance/cache.py
Draft review review/autogen_review.py

Other components

Clinic user

Connections 21

Clinic user → HTTP and chat API: submits query README.md
HTTP and chat API → Sequential agent crew: runs query api/main.py
HTTP and chat API → Masked event logging: logs request api/main.py
HTTP and chat API → Query budget: maps budget errors api/main.py
HTTP and chat API → Clinic policy documents: appends documents api/main.py
HTTP and chat API → Policy indexer: indexes additions api/main.py
Masked event logging → Request guardrails: masks phones api/logging_utils.py
Sequential agent crew → Request guardrails: checks requests agents/crew.py
Sequential agent crew → Query budget: checks budget README.md
Sequential agent crew → Generation cache: caches generation agents/crew.py
Sequential agent crew → Offline language model: uses model agents/crew.py
Sequential agent crew → Policy lookup tool: dispatches lookup README.md
Sequential agent crew → Appointment lookup: dispatches lookup README.md
Sequential agent crew → Draft review: reviews grounded drafts agents/crew.py
Sequential agent crew → Response schema: validates response agents/crew.py
Policy lookup tool → Similarity retrieval: retrieves evidence agents/tools.py
Similarity retrieval → Chroma collections: queries collection rag/retrieve.py
Policy indexer → Clinic policy documents: loads documents rag/index.py
Policy indexer → Chunk strategies: creates chunks rag/index.py
Policy indexer → Chroma collections: upserts embeddings rag/index.py
Appointment lookup → Appointment records: looks up records agents/tools.py

Drawn by GitDiagram on November 12, 2026 from the repository's file tree, README and source files. Read as Markdown ·