Architecture · LikeC4
pipegen-agent
Signal collection → complexity-first scoring → tiered outbound action list, plus per-account enrichment and an interactive account-intelligence bot
The whole picture
System landscape
The system in context, then opened up into its containers.
pipegen-agent — system landscape
index
pipegen-agent — containers (built vs deferred)
pipegenSystemSignal collection → complexity-first scoring → tiered outbound action list, plus per-account enrichment and an interactive account-intelligence bot
Risks & open questions
risks
Deferred & research work
plannedInside each box
Containers & components
Each part decomposed into the components that implement it. Every box links to its source in the interactive explorer.
Signal collection — components
signalsContainerGathers per-account signals from the open web and the internal data plane into per-account JSONL
Scoring & ranking pipeline — components
pipelineContainerPure-Python pipeline: build the candidate universe, gate on complexity, score, and rank. No learned model — a transparent baseline scorecard by design.
Deliverables — the three ranked views
deliverablesContainerThe artifacts handed to the sales team — three parallel ranked views over the same scored universe
Account enrichment — four-subagent fan-out + review gate
enrichmentContainerPer-account research, persona briefs, and outreach drafts for prioritized accounts, behind a human-review gate. PII-bearing artifacts stay gitignored.
Account-intelligence bot — components
botContainerChat-platform app that answers AE account questions, gated by AE-to-account ownership; dual-mode (local files in dev, managed cloud services in production)
How it runs
Walkthrough flows
Dynamic views — the narrative spine of the system, step by step.
Walkthrough — universe → gate → score → rank
scoringFlowThe core 5-stage pipeline. Three seeding sources merge, exclusions and the complexity gate fire before scoring, and the Strategy-B scorecard requires evidence to accumulate across components — no single signal can saturate the score.
Walkthrough — reading the internal data plane (no warehouse)
databotFlowThe project ingests nothing. A signal pull asks the data CLI a fixed natural-language question; the hosted LLM session generates and runs the query against the existing GTM data plane; contact PII is redacted before any raw response is written.
Walkthrough — account enrichment → review → AE distribution
enrichmentFlowFour specialist subagents fan out per account under a hard citation rule. PII-bearing artifacts stay gitignored; only the operator-approved, citation-stripped documents are mirrored to the shared drive for AEs.
Walkthrough — AE asks the account-intelligence bot
botFlowEvery answer is scoped to the AE’s owned accounts. The dispatcher routes to the right tool; data questions go through the same data CLI the pipeline uses.
Where it runs
Deployment
What runs where, and the process & data boundaries between the pieces.
Deployment — where each piece runs
deploymentExplore it live
The figures above are static exports. The interactive explorer lets you pan, zoom, follow relationships, and jump from any box to the source.