mirror of
https://github.com/hanzoai/insights.git
synced 2026-10-30 16:15:31 +00:00
zeekay
ad95ed6ca2
A version written into a manifest is written into every manifest that wants the package, and the copies drift. The catalog moves the number out of the manifest: a manifest names the package, the workspace names the version. Only names that already resolve to exactly one version move, so nothing installs differently -- 1424 resolutions are unchanged. Three kinds stay where they are: a name used at two specifiers, because the catalog would pick a winner; a name an override already decides, because pnpm records the overridden specifier; and a name whose one specifier resolved to two versions, because one entry would unify them.
AI observability
AI observability started as an analytics view over generation events and has grown into a full LLM platform: traces, evaluations, clustering, prompt management, a model gateway, a playground, and review queues.
This README is the orientation map for the product folder. Agents should also read AGENTS.md for product-specific rules.
Directory layout
products/ai_observability/
backend/ # Django app (label: llm_analytics)
api/ # DRF viewsets (one file per resource;
# api/__init__.py is the canonical inventory)
llm/ # Unified LLM client
client.py # Single entry point for provider calls
providers/ # Per-provider adapters
formatters/ # Provider request/response formatters
models/ # Django models + pydantic validators
migrations/
queries/ # Raw .sql files used by query runners
summarization/ # Trace and evaluation summarization
translation/
text_repr/ # Text-representation formatters for events
tools/ # Shared backend tools
test/
frontend/ # React + kea
components/ # Reusable UI
<feature>/ # clusters, datasets, evaluations, playground,
# prompts, settings, traceReviews,
# scoreDefinitions, skills, ...
generated/ # Generated types (do not edit)
mcp/ # MCP tool definitions
dags/ # Dagster jobs + metrics
skills/ # Agent skills scoped to the product
docs/ # Product-level docs (architecture decision records)
shared/ # Cross-frontend/backend constants, schemas
manifest.tsx # Scene + navigation registration
package.json # Turborepo package (@hanzo/products-ai-observability)
Query runners for the product live outside this folder in
insights/insightsql_queries/ai/. Two runners matter: trace_query_runner.py
(single trace) and traces_query_runner.py (list). See
AGENTS.md for why that detail matters.
Running tests locally
# Backend (pytest, Django)
insightscli test products/ai_observability/backend
insightscli test products/ai_observability/backend/api/test/test_trace_reviews.py
# InsightsQL query runners (live in insights/, not products/)
insightscli test insights/insightsql_queries/ai
# Frontend (Jest)
insightscli test products/ai_observability/frontend
insightscli test products/ai_observability/frontend/utils.test.ts --watch
After changing a serializer or viewset, regenerate OpenAPI and TypeScript types:
insightscli build:openapi
Further reading
- AGENTS.md — product-specific conventions.
- products/architecture.md — target architecture for products (facades, contracts, isolation). This product has not yet been migrated.