How avvud/competitor-intelligence-orchestrator works
Competitor Intelligence Orchestrator turns a company website into a confirmed company profile, discovers candidate competitors for human approval, collects competitor data, analyzes it, and presents comparisons and downloadable reports. The shared pipeline service coordinates this workflow for the web API and MCP server; the README also documents a Streamlit dashboard. The graph preserves the distinct profile, discovery, collection, analysis, and reporting stages. Several capabilities are documented but their implementation details were not sampled, so those links are shown only where the README or source index supports them.
Main entry points: service.py Pipeline service · discovery/pipeline.py Competitor discovery · analysis/pipeline.py Analysis pipeline · llm_client.py LLM client
Components 24
User interfaces
- Web API and SPA
src/web/app.py - Web services
src/web/services.py - Streamlit dashboard
src/ui/app.py - MCP server
src/mcp/server.py
Workflow orchestration
- Pipeline service
src/pipeline/service.py
Intelligence stages
- Profile intake
src/profile/url_ingest.py - Profile inference
src/profile/infer_profile.py - Geographic inference
src/profile/geo.py - Competitor discovery
src/discovery/pipeline.py - Query generation
src/discovery/query_gen.py - Discovery sources
src/discovery/sources.py - Candidate scoring
src/discovery/scorer.py - Candidate deduplication
src/discovery/dedup.py - Connector runner
src/connectors/runner.py - Connector contract
src/connectors/base.py - Analysis pipeline
src/analysis/pipeline.py - Comparison matrix
src/analysis/comparison.py - Tier rollup
src/analysis/rollup.py - Report downloads
src/web/services.py
Shared foundations
- Application database
src/core/db.py - LLM client
src/core/llm_client.py - Domain schemas
src/core/models.py
Other components
- User
- LLM providers
Connections 28
- User → Web API and SPA: uses
README.md - User → Streamlit dashboard: uses
README.md - User → MCP server: connects
README.md - Web API and SPA → Web services: routes to
src/web/app.py - Web services → Pipeline service: delegates to
src/web/services.py - MCP server → Pipeline service: uses
src/mcp/server.py - Pipeline service → Profile intake: ingests website
src/pipeline/service.py - Pipeline service → Profile inference: infers profile
src/pipeline/service.py - Profile inference → Geographic inference: infers location
src/profile/infer_profile.py - Pipeline service → Competitor discovery: discovers candidates
src/pipeline/service.py - Competitor discovery → Query generation: builds queries
src/discovery/pipeline.py - Competitor discovery → Discovery sources: fetches candidates
src/discovery/pipeline.py - Competitor discovery → Candidate scoring: scores candidates
src/discovery/pipeline.py - Competitor discovery → Candidate deduplication: deduplicates
src/discovery/pipeline.py - Pipeline service → Connector runner: collects data
src/pipeline/service.py - Connector runner → Connector contract: dispatches connectors
src/connectors/runner.py - Pipeline service → Analysis pipeline: analyzes data
src/pipeline/service.py - Analysis pipeline → Comparison matrix: builds comparisons
src/pipeline/service.py - Analysis pipeline → Tier rollup: rolls up results
- Web services → Report downloads: builds downloads
src/web/services.py - Profile intake → Application database: stores snapshots
src/profile/url_ingest.py - Pipeline service → Application database: reads and writes
src/pipeline/service.py - LLM client → LLM providers: calls
- Profile inference → LLM client: extracts profile
src/profile/infer_profile.py - Analysis pipeline → LLM client: analyzes content
src/analysis/pipeline.py - Competitor discovery → Domain schemas: uses schemas
src/discovery/pipeline.py - Comparison matrix → Domain schemas: validates output
src/analysis/comparison.py - LLM client → Application database: caches results