How mmansii/builderr-signalpost works
Signalpost is a Norwegian-company research agent that turns organisation-number batches into auditable company profiles and terminal JSONL envelopes. Its main workflow selects or accepts companies, anchors them in the Brønnøysund registry, gathers official and website evidence, checks exact-entity identity, and emits results with provenance and run telemetry. Separate capabilities cover replaying profile changes, screening and evidence-bounded questions, and collecting external-footprint signals. The graph reflects documented workflow and indexed imports; some connector implementations and runtime wiring were not sampled, so uncertain links are omitted.
Main entry points: batch.py Competition batch · official.py Official records · website.py Company website · evidence.py Evidence records
Components 22
Input and registry
- Batch input
src/norway_company_agent/batch.py - Company sampling
src/norway_company_agent/sampling.py
Evidence gathering
- BRREG registry
src/norway_company_agent/official.py - Official records
src/norway_company_agent/official.py - HTTP fetching
src/norway_company_agent/http.py - Company website
src/norway_company_agent/website.py - Site discovery
src/norway_company_agent/discovery.py - Crawl extraction
src/norway_company_agent/crawl_events.py - Resumable crawler
src/norway_company_agent/scrapy_crawler.py
Identity and evidence
- Entity identity gate
src/norway_company_agent/identity.py - Evidence records
src/norway_company_agent/evidence.py
Profiles and analysis
- Competition batch
src/norway_company_agent/batch.py - Refresh replay
src/norway_company_agent/refresh.py - Saved snapshots
src/norway_company_agent/snapshots.py - Research and screening
src/norway_company_agent/research.py - Research workspace
src/norway_company_agent/workspace.py
External signals
- Footprint validation
src/norway_company_agent/external_footprint.py - Coverage evaluation
src/norway_company_agent/external_control.py - Sentiment analysis
src/norway_company_agent/sentiment.py - External connectors
scripts
Other components
- Research operator
- Company data sources
Connections 24
- Research operator → Company sampling: selects sample
- Research operator → Batch input: submits batch
- Batch input → Competition batch: provides companies
src/norway_company_agent/batch.py - Company sampling → Batch input: produces batch
src/norway_company_agent/batch.py - Competition batch → BRREG registry: anchors identity
src/norway_company_agent/batch.py - BRREG registry → Official records: fetches records
src/norway_company_agent/official.py - Official records → HTTP fetching: fetches JSON
src/norway_company_agent/official.py - Competition batch → Company website: visits listed site
- Company website → Site discovery: finds site candidates
src/norway_company_agent/discovery.py - Resumable crawler → Crawl extraction: emits page events
src/norway_company_agent/scrapy_crawler.py - Crawl extraction → Evidence records: records website evidence
src/norway_company_agent/crawl_events.py - Competition batch → Entity identity gate: checks entity match
- Entity identity gate → Evidence records: gates publication
src/norway_company_agent/identity.py - Competition batch → Evidence records: records provenance
src/norway_company_agent/batch.py - Competition batch → Research operator: emits envelopes
- Refresh replay → Saved snapshots: replays snapshots
src/norway_company_agent/snapshots.py - Research and screening → Research workspace: uses workspace
scripts/evaluate_research_agent.py - Research operator → Research and screening: asks or screens
- External connectors → Footprint validation: supplies observations
scripts/evaluate_external_footprint.py - Coverage evaluation → Footprint validation: scores footprint
src/norway_company_agent/external_control.py - Sentiment analysis → Footprint validation: enriches observations
- Company data sources → Official records: provides records
- Company data sources → Company website: hosts company site
- Company data sources → External connectors: provides public signals