Conversation
Implements the full ProdtrackProviderBase surface over ftrack, alongside the existing ShotGrid and mock providers. Select with PRODTRACK_PROVIDER=ftrack. Three decisions shape the implementation: Entity ids. ftrack keys entities by UUID while DNA ids are int throughout (EntityBase, route params, Mongo bookkeeping, frontend types). The provider assigns a stable int per UUID and persists the pairing in ftrack_id_map.py, so a reverse lookup resolves on any instance after any restart. The map must outlive the process: draft notes, transcripts and stored segments all reference these ids. Playlists. ftrack has two candidates — AssetVersionList (Lists, the default) and ReviewSession (Client Reviews). Both are PlaylistAdapter subclasses, and playlist_adapter() resolves the type per call from FTRACK_PLAYLIST_ENTITY, so it can be swapped for a user setting later without touching anything built on top. Query shape. Projections are flat: scalar columns and foreign keys only, never a dotted path. Deep projections make the ftrack server build the joins, which is slower than fetching each layer separately. _hydrate_versions walks versions -> assets -> contexts -> object types and stitches the result, one batched query per layer, so a playlist costs a constant number of queries regardless of size. Two tests reject a dotted path in any projection constant or playlist-load query. Also adds transcript_entity_type() to the provider base so main.py no longer reads ShotGrid's env var directly, and a thumbnail proxy endpoint because ftrack's own thumbnail URL carries the API key in its query string. Known gaps vs ShotGrid (note subject in metadata, cc merged into recipients, updated_at unavailable, opt-in media paths) are documented in backend/README.md. Not yet run against a live ftrack server, and the repo test suite has not run against it either — see the handoff doc for what that leaves unverified. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
Provides the toolchain that was missing on the host and blocked running the backend tests outside Docker: python311 (matching backend/Dockerfile), uv, nodejs_22, mongodb-ce, black, isort and docker-compose. Also packages the two prodtrack SDKs that are absent from nixpkgs, which pulls in two more: shotgun_api3 3.9.2 github shotgunsoftware/python-api ftrack-python-api 3.0.6 github ftrackhq/ftrack-python-api clique 1.6.1 PyPI (not tagged on GitHub past 1.5.0) arrow 0.17.0 github arrow-py/arrow arrow is pinned below 1.0 deliberately. ftrack calls arrow.now().replace(tzinfo="utc") in entity/user.py, and arrow 1.0 removed tzinfo from .replace(); nixpkgs ships 1.4, which would raise on a path get_versions_for_playlist reaches. pyparsing and websocket-client are relaxed instead, since they are only used by the event hub and the provider connects with auto_connect_event_hub=False. ftrack needs build patching: upstream uses poetry-dynamic-versioning, which reads the version from a git tag, so a source tarball would build as 0.1.0. The derivation pins the version, swaps to plain poetry-core, and drops sphinx-notfound-page (declared as a runtime dep but docs-only). The shell does not install backend/requirements.txt automatically — it prints the uv commands instead, so entering it is fast and works offline. Tests still run in Docker. Adds backend/docs/ftrack-provider-handoff.md covering the provider's design, its verified and unverified assumptions, and this flake. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
requirements.txt pinned only top-level packages, so transitive dependencies floated per install and anthropic was a range. Declare the dependencies in backend/pyproject.toml, lock the full tree in backend/uv.lock, and build the nix dev shell's Python environment from it with uv2nix. requirements.txt is now exported from the lock, so the dev shell, Docker and CI install the same set. This replaces the hand-built shotgun_api3, ftrack-python-api, arrow and clique derivations; uv resolves them from PyPI and honours ftrack's pins instead of relaxing them. Also allow the unfree mongodb-ce (SSPL) through allowUnfreePredicate, without which the flake failed to evaluate. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
NOTE_PROJECTION fetches only author_id, so reading note["author"] made ftrack auto-populate the link with an extra query per note. The author is now fetched explicitly by id when links are resolved, and passed into _note_from rather than read off the note. A freshly created note keeps its locally populated author. Adds tests for single-entity reads, hydration of each entity type, the thumbnail fetch, and the /api/ftrack-thumbnails endpoint. Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
Adds two apps that run mongo, the backend and the frontend on the host
from the working tree, so uncommitted changes are what gets exercised:
nix run .#dev all three under process-compose, backend reloading,
mock tracker and no auth by default
nix run .#smoke the same headless on spare ports and a throwaway
database; checks health, projects -> playlists ->
versions through the tracker, a draft note round
trip through storage, that the frontend serves and
compiles its entry, and CORS; then stops everything
Settings live in nix/stack-env.sh. Exported variables or .dna-dev/env
override them, e.g. PRODTRACK_PROVIDER=ftrack with its credentials.
Both apps refuse to start on a busy port instead of crash-looping.
The dev shell now lists the backend checks as CI runs them instead of
pointing at Docker, marks its prompt with (dna), and adds
process-compose and jq.
Co-Authored-By: Claude Opus 5.5 (1M context) <noreply@anthropic.com>
|
8 tasks
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Summary
Describe your change.
Testing
How I Tested
Describe exact steps used to test.