Add CLAUDE.md for Claude Code guidance
Co-Authored-By: Claude Sonnet 4.6 (1M context) <noreply@anthropic.com>
This commit is contained in:
parent
546242e11a
commit
dfc76a96d8
99
CLAUDE.md
Normal file
99
CLAUDE.md
Normal file
@ -0,0 +1,99 @@
|
|||||||
|
# CLAUDE.md
|
||||||
|
|
||||||
|
This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.
|
||||||
|
|
||||||
|
## Running the Server
|
||||||
|
|
||||||
|
```bash
|
||||||
|
pipekit serve # default host/port from config.yaml
|
||||||
|
pipekit serve --host 0.0.0.0 --port 8080 --reload # dev mode with auto-reload
|
||||||
|
```
|
||||||
|
|
||||||
|
The user runs the server themselves in their own terminal — do not nohup or background it.
|
||||||
|
|
||||||
|
## Other CLI Commands
|
||||||
|
|
||||||
|
```bash
|
||||||
|
pipekit init # create/upgrade SQLite schema
|
||||||
|
pipekit doctor # health check (config, jrunner, DB)
|
||||||
|
pipekit run <module_name> # run a module synchronously (manual test)
|
||||||
|
pipekit set-password <username> # set HTTP Basic Auth credentials
|
||||||
|
pipekit secrets set KEY [VALUE] # add/update a secret (prompts if value omitted)
|
||||||
|
pipekit drivers list # show registered driver kinds
|
||||||
|
./deploy.sh # idempotent: venv, deps, launcher, driver registration
|
||||||
|
```
|
||||||
|
|
||||||
|
## Architecture
|
||||||
|
|
||||||
|
Pipekit is a database sync tool. A **module** defines a source query → dest table sync job. The engine runs a module by: resolving watermarks → materializing source SQL → staging data via jrunner → merging into dest.
|
||||||
|
|
||||||
|
**Layers (bottom to top):**
|
||||||
|
|
||||||
|
1. **SQLite** (`pipekit.db`) — single file, all state
|
||||||
|
2. **`repo.py`** — all CRUD for every table; ~1,900 LOC, the only layer that touches the DB
|
||||||
|
3. **`engine/`** — orchestrates a module run: lock acquisition, watermark resolution, jrunner calls, merge SQL, post-run hooks, run_log write, lock release
|
||||||
|
4. **jrunner** — external Java CLI; handles all JDBC access. Python never talks to remote DBs directly; it shells out to jrunner
|
||||||
|
5. **`api/`** — FastAPI REST endpoints under `/api/*`, HTTP Basic Auth (except `/health`)
|
||||||
|
6. **`web/`** — HTML pages (Jinja2 templates) at `/`; HTMX + Alpine.js for interactivity
|
||||||
|
|
||||||
|
## Key Data Model
|
||||||
|
|
||||||
|
- **driver** — JDBC driver registration (kind, jar, class, url_template)
|
||||||
|
- **connection** — named DB connection (jdbc_url, username, password as `$ENV_VAR` reference)
|
||||||
|
- **module** — sync job (source/dest connection, source query, merge strategy, merge key, enabled, running lock)
|
||||||
|
- **watermark** — named placeholder with resolver SQL; first column of first row used as opaque string; replaces `{watermark_name}` in source query
|
||||||
|
- **hook** — post-merge SQL (run_order, run_on: success/failure/always)
|
||||||
|
- **run_log** — immutable history record with resolved SQL, merge SQL, watermark values, stdout/stderr, timing
|
||||||
|
|
||||||
|
## Engine Flow
|
||||||
|
|
||||||
|
```
|
||||||
|
run_module(module_id)
|
||||||
|
→ atomic UPDATE module SET running=1 WHERE running=0 # fail if already locked
|
||||||
|
→ for each watermark: run resolver SQL via jrunner → capture first cell
|
||||||
|
→ materialize source query (simple string replace {name} → value)
|
||||||
|
→ jrunner create staging table (pipekit_staging.{module_name})
|
||||||
|
→ jrunner migrate source → staging
|
||||||
|
→ build merge SQL (engine/merge.py: full=truncate+insert, incremental=delete by key+insert, append=insert)
|
||||||
|
→ run merge SQL via jrunner
|
||||||
|
→ run hooks in order
|
||||||
|
→ write run_log entry
|
||||||
|
→ UPDATE module SET running=0 (in finally)
|
||||||
|
```
|
||||||
|
|
||||||
|
## Credentials Pattern
|
||||||
|
|
||||||
|
Passwords in connections are stored as `$VAR_NAME` references. At run time they are resolved from `/etc/pipekit/secrets.env` (override with `PIPEKIT_SECRETS` env var). Config path override: `PIPEKIT_CONFIG`.
|
||||||
|
|
||||||
|
## Driver Abstraction
|
||||||
|
|
||||||
|
Each driver in `pipekit/drivers/` inherits from `base.py::Driver` and implements: `browse_fields`, `list_tables`, `list_schemas`, `get_columns`, `map_type`, `default_expression`, `quote_identifier`. The wizard UI calls `/api/introspect/*` which dispatches to the appropriate driver.
|
||||||
|
|
||||||
|
## Module Columns
|
||||||
|
|
||||||
|
Modules store their column mapping as `columns_json` — a JSON list of dicts with keys `source_name`, `source_type`, `dest_name`, `dest_type`. The engine uses this to build the staging CREATE TABLE and the merge INSERT column lists.
|
||||||
|
|
||||||
|
## Merge Key
|
||||||
|
|
||||||
|
`merge_key` is stored as a comma-separated string (e.g., `"col1, col2"`). The engine parses it and generates a multi-column DELETE predicate for incremental strategy.
|
||||||
|
|
||||||
|
## Staging Table
|
||||||
|
|
||||||
|
Recreated on every run as `pipekit_staging.{module_name}` (DROP + CREATE, not IF NOT EXISTS). Ephemeral — exists only during the run.
|
||||||
|
|
||||||
|
## API vs. Web
|
||||||
|
|
||||||
|
- `/api/*` — JSON REST, HTTP Basic Auth, consumed by HTMX fragments and external callers
|
||||||
|
- `/` and other bare paths — full HTML pages (Jinja2), no auth currently
|
||||||
|
- `POST /modules/{id}/run` returns `{run_id}` immediately; run is async; poll `/runs/{id}` for status
|
||||||
|
|
||||||
|
## Tech Stack
|
||||||
|
|
||||||
|
- Python 3.10+, FastAPI, Uvicorn, Jinja2, PyYAML, SQLite3 (stdlib)
|
||||||
|
- `python-multipart` required for HTML form POSTs (not auto-installed as a FastAPI transitive dep)
|
||||||
|
- Frontend: HTMX + Alpine.js (CDN), no build step
|
||||||
|
- jrunner: separate Java tool, must be on PATH
|
||||||
|
|
||||||
|
## Full Spec
|
||||||
|
|
||||||
|
`/opt/pipekit/SPEC.md` is the authoritative design document. Read it for deep rationale on any architectural decision.
|
||||||
Loading…
Reference in New Issue
Block a user