Skip to content

Epic: Prophet Understand / Repo Intelligence v0 #407

@mdheller

Description

@mdheller

Purpose

Create the SocioProphet repo-understanding layer: a governed, searchable, agent-usable codebase and knowledge-base graph artifact for every important repository.

This is the strategic absorption path from the current open-source "Understand Anything" style product pattern. We should borrow the product lesson — interactive graph, guided architecture tours, semantic search, diff impact analysis, and multi-agent packaging — but implement it as a governed SocioProphet substrate rather than vendoring an ungoverned local JSON/dashboard tool.

Product thesis

Every repo should carry a machine-readable self-model that humans and agents can trust:

  • structural graph of files, modules, services, contracts, endpoints, schemas, docs, domains, and flows
  • human-readable summaries for each important node
  • guided tours ordered by dependency and risk
  • semantic/fuzzy/graph search over repo contents and concepts
  • PR diff impact analysis with affected nodes, edges, tests, docs, and policies
  • provenance receipts tying graph facts to commit, parser, agent, source hash, and confidence
  • policy gates that detect stale, unsigned, invalid, or drifted repo-intelligence artifacts

Core artifact

Define a versioned prophet-understanding.json / sociograph.json artifact family with at least:

  • schema_version
  • repo, commit, default_branch, generated_at
  • generator and parser versions
  • agent_identity / future DID binding placeholder
  • source_hashes
  • nodes[]
  • edges[]
  • summaries[]
  • tours[]
  • diff_impact[]
  • policy_status
  • provenance_receipts[]
  • validation_results[]

Cross-repo ownership

  • smart-tree: structural scanner and local graph emitter
  • lampstand: local index/search over repo graph artifacts
  • sherlock-search: hybrid lexical/vector/graph query over repo intelligence
  • ontogenesis: ontology, JSON-LD, SHACL, Avro/TriTRPC contract generation
  • agent-registry: multi-agent skill manifests from one registry contract
  • agentplane: graph-aware dispatch, bounded repair, and impact-aware agent work orders
  • policy-fabric: validation gates for graph freshness, provenance, signatures, and PR impact risk
  • delivery-excellence: metrics for freshness, coverage, graph drift, and agent repair quality
  • socioprophet: /repo-map or /understand UI workbench

v0 acceptance criteria

  • A first formal artifact contract exists and validates in CI.
  • At least one fixture repo graph is committed as an example.
  • A scanner can emit a deterministic graph from one repo.
  • Search can answer at least: "what owns this file?", "what depends on this contract?", "what changes in this PR matter?"
  • UI workbench has a minimal graph + node detail + evidence panel path.
  • Policy gates detect stale/missing/invalid artifacts without blocking unrelated work by default.
  • Metrics expose graph freshness, coverage, drift, and PR impact radius.

Non-goals for v0

  • No autonomous code mutation from graph output.
  • No blind vendoring of Lum1104/Understand-Anything.
  • No unreviewed post-commit hooks in sensitive repos.
  • No claims of semantic certainty without evidence/confidence.

Implementation stance

Build our own governed substrate. Use external projects as benchmarks and UX references, not as the trust root.

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type
    No fields configured for issues without a type.

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions