folio-notion 0.1.0

Orchestrator: PDFs → NotebookLM → synthesis → disk + Notion

Folio Notion

User story: configure once → add PDFs → stay logged into NotebookLM → run → local export + Notion.

This repo is a single product: an orchestrator that owns the sequence and glue between tools you already use (files, NotebookLM, disk, Notion). It does not replace those tools; it drives them in order and passes data between steps.

For technical readers: this is a research / ingestion pipeline with orchestration at the center.

Layout

See the tree below; each top-level folder has a short README.md where it helps.

folio-notion/
├── README.md
├── pyproject.toml
├── .env.example
├── .gitignore
├── config/                 # checked-in templates; user config lives outside or via env
├── docs/
│   ├── user-workflow.md    # journey + milestones (OSS-friendly)
│   └── roadmap.md
├── src/folio_notion/       # application package
│   ├── cli.py              # “Run pipeline” entrypoint
│   ├── pipeline.py         # sequence: configure → … → Notion
│   ├── steps/              # one module per pipeline stage
│   └── integrations/       # NotebookLM, Notion API, filesystem
├── tests/
├── scripts/                # optional one-off maintenance / dev helpers
└── var/                    # default local scratch (gitignored); exports, caches

Quick start

Install from PyPI (after you publish the package):

pip install folio-notion
# or: pip install folio_notion   # same package (normalized name)

# NotebookLM + browser flow:
pip install "folio-notion[notebooklm]"
playwright install chromium

Install from a git clone (development):

cd folio-notion
python -m venv .venv && source .venv/bin/activate   # Windows: .venv\Scripts\activate
pip install -e .

Run pipeline (PDFs → NotebookLM chat → var/exports + Notion page):

pip install -e ".[notebooklm]"   # NotebookLM API + browser login
playwright install chromium
# Put PDFs in var/inbox (or set input_dir in config/config.yaml)
fn run                            # same as: fn run -C .
fn run -C /path/to/folio-notion
python -m folio_notion

Run modes

• fn run --dry-run — load config and verify NotebookLM storage (+ Notion, unless you combine with --skip-notion). Does not ingest PDFs or call NotebookLM.
• fn run --skip-notion — NotebookLM synthesis and local markdown under export_dir; skips Notion API checks and page creation (useful for end-to-end LM testing without publishing).
• PDFs — by default Folio looks in input_dir (often var/inbox). If it’s empty and your terminal is interactive, it prompts for file or folder path(s) (comma-separated). Non-interactive / CI: fn run --pdf /path/to/file.pdf or --pdf /path/to/folder (repeat --pdf for several paths).

Structured logs: set FOLIO_LOG_LEVEL to DEBUG or INFO (default).

Status — see whether Notion / NotebookLM look connected (uses a quick API check; needs network):

fn status
fn status -C /path/to/folio-notion

Requires .env with NOTION_TOKEN, NOTION_PARENT_PAGE_ID, and FOLIO_NOTEBOOKLM_STORAGE (from the connect commands). Config: config/config.yaml (merged with config/defaults.example.yaml).

Connect Notion — verify your internal integration secret with the API, optionally check a parent page id, optionally write .env:

fn connect notion
# alias:
fn notion connect

Flags: --token SECRET, --parent PAGE_ID, --save, --no-save, -C /path/to/project (where .env lives).

Interactive prompts use prompt-toolkit so arrow keys and normal line editing work (after pip install -e .). Parent may be a Notion page or database; database URLs are verified via the databases API.

Interactive parent step: bad input keeps prompting (a short guide appears every 3 failures). Enter skips parent; quit / exit / q leaves parent empty but continues to the save question.

Do not paste lines from pyproject.toml (like [project.scripts]) into the shell; those belong only in the file.

Connect NotebookLM (browser → storage → httpx)

Architecture: interactive Playwright session once → save storage_state.json → httpx loads cookies and fetches NotebookLM page tokens (SNlM0e, FdrFJe) for future RPC-style calls. Cookie parsing patterns are aligned with notebooklm-py; paths default to ~/.folio-notion/notebooklm/ (override with NOTEBOOKLM_HOME / FOLIO_NOTEBOOKLM_HOME).

pip install -e ".[notebooklm]"
playwright install chromium   # if not already installed
fn connect notebooklm
# alias:
fn notebooklm connect

• Chromium uses stealth-ish flags (AutomationControlled, no --enable-automation banner); persistent profile under browser_profile/.
• Session file is storage_state.json (mode 600 where supported). Optional: NOTEBOOKLM_AUTH_JSON for CI (same shape as Playwright storage).
• Set FOLIO_PLAYWRIGHT_AUTO_INSTALL=0 to skip automatic playwright install chromium.
• Flags: --storage PATH, --no-verify, --save / --no-save, -C project root (for .env).

Configuration

Paths, prompts, Notion parent page, optional notebook id — configure once via config/ templates and environment variables (see .env.example). After fn connect notion, NOTION_TOKEN and optionally NOTION_PARENT_PAGE_ID can live in .env. After fn connect notebooklm, FOLIO_NOTEBOOKLM_STORAGE can point at your saved storage_state.json.