Skip to content

[SEED-071] NotebookLM Foundation - RAG-Backed Semantic Memory for t27 #305

Closed

Description

@gHashTag

SEED: NotebookLM Integration for t27

Ring Assignment: Ring-071
Status: Pending
Created: 2026-04-07

Summary

Integrate Google NotebookLM as RAG-backed semantic memory for t27 to solve session amnesia. Each Claude Code session can upload wrap-up summaries to NotebookLM, enabling semantic search across all sessions with minimal token costs.

Hypothesis

NotebookLM integration provides "infinite memory" with:

  • Persistent storage of session wrap-ups
  • Semantic search via RAG
  • Minimal token costs (search results retrieved, not regenerated)
  • Natural language query interface

Research Outcomes

API Strategy Decision: Use notebooklm-py SDK (v0.3.4)

  • Reverse-engineered Protobuf RPCs
  • Cookie-based authentication via Playwright storage state
  • Full SDK with 5.6k+ GitHub stars
  • No official consumer API available

Feasibility: Verified

  • SDK installation and API exploration complete
  • Authentication flow documented
  • Async nature requires wrapper functions
  • Storage state: ~/.notebooklm/storage_state.json

Decomposition Plan

27 atomic tasks across 5 phases:

  • Phase 0 (T01-T04): Research & Decision - COMPLETED
  • Phase 1 (T05-T09): Specification - .t27 spec creation
  • Phase 2 (T10-T17): Core Implementation - Python modules
  • Phase 3 (T18-T22): Session Integration - Wrap-up skills
  • Phase 4-5 (T23-T27): Testing & Deployment

See full plan in: docs/research/notebooklm-integration-plan.md

Acceptance Criteria

  1. specs/memory/notebooklm.t27 created and validated
  2. contrib/backend/notebooklm/ implemented with all modules
  3. Wrap-up skill uploads session summaries to NotebookLM
  4. Unit tests pass with >80% coverage
  5. E2E test: Create notebook → Upload summary → Query → Verify
  6. Benchmarks meet criteria (OAuth <2000ms, Notebook create <3000ms, etc.)
  7. Seal applied and episode recorded

Constitutional Compliance

  • L1 TRACEABILITY: This issue tracks all commits
  • L2 GENERATION: Specs are SSOT; generated code never hand-edited
  • L3 PURITY: ASCII-only English
  • L4 TESTABILITY: test/invariant/bench in every spec
  • L7 UNITY: Use tri/t27c, no new shell scripts

Risks & Mitigations

Risk Probability Mitigation
notebooklm-py SDK breaks Medium Fallback to Playwright automation
Google blocks cookie auth Medium Migrate to Enterprise API
Rate limiting High Exponential backoff + local cache

References

  • Feasibility report: .trinity/audit/notebooklm-feasibility.md
  • Inventory: .trinity/audit/inventory.json
  • AKASHIC LOG: .trinity/events/akashic-log.jsonl
  • Episodes: .trinity/experience/episodes.jsonl

Activity

  1. gHashTag commented on Apr 8, 2026

    @gHashTag
    CollaboratorAuthor

    📓 NotebookLM Notebook created

    Notebook ID: 4a801a0b-de83-4bd2-b05c-e82a2f49f489
    URL: https://notebooklm.google.com/notebook/4a801a0b-de83-4bd2-b05c-e82a2f49f489

  2. gHashTag commented on Apr 29, 2026

    @gHashTag
    CollaboratorAuthor

    Implementation complete:

    • Spec: specs/memory/notebooklm.t27 (485 lines, 11 tests, 8 invariants, 4 benchmarks — L4 compliant)
    • Backend: contrib/backend/notebooklm/ (12 modules, ~4000 lines)
    • Tests: contrib/backend/notebooklm/tests/ (8 test files)
    • Auth, CRUD, wrapup, sync all implemented
    • Acceptance criteria 1-5 satisfied
  3. added 12 commits that reference this issue on May 5, 2026
    1900b15
    3443211
    af3e4c7
    9dfbeae
    c0b5f5b
    cd1c5af
    18d6ceb
    94371a6
    798e5f2
    3640832
    4f4265d
    1cbc0b0
  4. 17 remaining items

  5. added 15 commits that reference this issue on Jun 1, 2026
    82eb8f7
    58777de
    f4d5d41
    fdc6b7a
    9cd1cc4
    4d61341
    d57ecea
    43f9e40
    3e9f851
    a513183
    d600d5e
    3af2ec0
    f89eade
    4798a4b
    597dc5a
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    No labels
    No labels

    Type

    No type

    Projects

    No projects

      Milestone

      No milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions