Project and delivery
Objective
Deliver an independently installable Python 3.11+ RAG runtime that owns generic immutable document/query/result models, injected embedding/vector/lexical ports, rebuildable derived projection, and deterministic semantic/hybrid retrieval without depending on story-writing-agents or any application-specific storage.
Integration criteria
- The distribution/import identities are
generic-rag and generic_rag.
- Runtime dependencies remain empty; mandatory tests use deterministic fakes and no network.
- The package imports no
story_writing_agents, SQLite/MySQL, vendor SDK, credential, workspace, session, conversation, agent, or citation concept.
- Projection state is derived, versioned, checkpointed, rebuildable, and deletable through opaque corpus/document revisions.
- Hybrid fusion is rank-based and deterministic rather than comparing provider-specific raw scores.
- Importing the package performs no I/O, authentication, subprocess, thread, or application startup.
- The package has its own policies, module index, build, tests, documentation, releases, Issues, and PRs.
- Story integration consumes only the reviewed public package API through a pinned Git submodule commit.
Child inventory and order
Known child leaves will cover:
- Repository/package bootstrap plus public models and injected ports.
- Revision-aware projection, checkpoint, rebuild, and deletion.
- Semantic retrieval, deterministic hybrid fusion, composition, verification, and documentation.
The first leaf establishes the contracts. Projection and retrieval follow through native dependencies.
Non-goals
Story-specific events/citations, authoritative history ownership, SQLite/MySQL adapters, vendor/provider adapters, dynamic loading, credentials, implicit networking, raw-score cross-provider ordering, or a CLI.
Risks
Premature application coupling, unstable public models, provider-score nondeterminism, hidden resource ownership, and submodule/release drift.
Gates
github_planning_gate = pending until all descendants, relationships, Project metadata, milestone assignments, and the first Development-linked leaf branch are verified.
modularity_gate = required. This repository requires its own Python policy and PYTHON_MODULE_INDEX.md before structural production work.
Definition of complete
Project and delivery
mainat33ef381Objective
Deliver an independently installable Python 3.11+ RAG runtime that owns generic immutable document/query/result models, injected embedding/vector/lexical ports, rebuildable derived projection, and deterministic semantic/hybrid retrieval without depending on story-writing-agents or any application-specific storage.
Integration criteria
generic-ragandgeneric_rag.story_writing_agents, SQLite/MySQL, vendor SDK, credential, workspace, session, conversation, agent, or citation concept.Child inventory and order
Known child leaves will cover:
The first leaf establishes the contracts. Projection and retrieval follow through native dependencies.
Non-goals
Story-specific events/citations, authoritative history ownership, SQLite/MySQL adapters, vendor/provider adapters, dynamic loading, credentials, implicit networking, raw-score cross-provider ordering, or a CLI.
Risks
Premature application coupling, unstable public models, provider-score nondeterminism, hidden resource ownership, and submodule/release drift.
Gates
github_planning_gate = pendinguntil all descendants, relationships, Project metadata, milestone assignments, and the first Development-linked leaf branch are verified.modularity_gate = required. This repository requires its own Python policy andPYTHON_MODULE_INDEX.mdbefore structural production work.Definition of complete