Input is user and assistant text from the opted-in Pi JSONL artifact.
Review the work
agents do.
Register insight services that you deploy yourself. Apply policies globally or by tenant and project. Connect each result to the repository, session, source items, tags, and run that produced it. You manage this through your own gateway.
Manage policies for each insight service.
The gateway parses canonical Pi sessions. It matches every active global, tenant, and project configuration. It calls the selected stateless gRPC processor. The gateway owns durable identity and provenance. The service owns the insight logic.
The gateway bounds the input, removes thinking and tool records, and resolves tags and configurations.
The processor returns summaries, tagged results, source item IDs, usage, warnings, and proposed tags.
The gateway stores the run and provenance. It appends accepted results to project knowledge.
Register the endpoint
Provide a name, gRPC target, TLS or trusted-network plaintext, server name, timeout, active state, and a live health test.
Scope the policy
Attach a processor, instructions, tag behavior, summaries, input bounds, and a revision to global, tenant, or project scope.
Control the tag vocabulary
Supply active tag names, descriptions, and usage counts to every service. Choose whether a configuration may propose new tags.
Review each run
View configuration revision, project, input hash, status, session summary, insight count, usage, warnings, errors, and timing.
Use one project for several checkout paths.
The plugin reports Git remotes. The gateway normalizes SSH and HTTPS forms into a tenant-scoped project identity. Activity from different machines converges. Unrelated repositories that merely share a folder name do not merge.
git@github.com:latchlabs-dev/yapp.git https://github.com/latchlabs-dev/yapp /Users/team-a/code/yapp See four views of what the agent learned.
Raw session evidence, generated insight, prompt quality, and curated memory serve different jobs. The gateway keeps those layers distinct.
Review lessons from agent sessions
Agent observations and the distillation pass produce durable decisions, mistakes, lessons, and context. Teams can review them independently of a raw session.
See where a request caused rework
Project views keep prompt reflections separate from operational memory. Request quality stays visible without mixing into the project runbook.
Compact repeated lessons
Repeated lessons can be compacted into a focused project brief. Retrieval and injection remain policy-controlled and can be disabled centrally.
See where an insight came from
Entries preserve their source type, conversation, and user attribution when available. Teams can validate the evidence before acting on it.
Use the stable repository remote as project identity. Folder names are display hints only.
recorded by agent · conversation 7ba…When the knowledge store is unavailable, do not report an empty response as zero insights.
generated by distiller · user attributedUsage attribution arrives from accepted proxy requests even when a canonical session upload is disabled.
recorded by agent · source linkedMake insights auditable.
The project view reports pending, skipped, failed, and complete distillation work. If the knowledge store is unavailable, the gateway says so instead of displaying a misleading zero.
- Distillation is opt-in and project allowlisted
- Raw session upload remains a separate client opt-in
- Knowledge entries preserve available attribution
- Curated memory can be inspected before it is reused
See which client and processor are connected.
gRPC health tests, version headers, session metadata, run failures, and aggregated extension reports make both sides of the insight pipeline visible.