feat(bundles): metapackage split (Phase A) — engine-only lfx, lfx-bundles long tail, 5 graduated partner packages - #13563
Conversation
… tier Foundation for the bundle metapackage split (1.11). Adds a third @official-slot discovery source: a distribution declaring the [project.entry-points."lfx.bundles"] group ships a package whose immediate subdirectories are each a manifest-less bundle, folder-walked and registered at @official with no extension.json (the langchain-community model). - new loader/_bundles_root.py::load_lfx_bundles_extensions, mirroring discover_inline_bundles but sourcing roots from the lfx.bundles entry-point group via find_spec and reusing _load_bundle_directory - discovery precedence becomes installed > seed > lfx_bundles > dev > inline so a manifest-shipping lfx-<provider> always shadows the same-named provider in the metapackage (lets a provider graduate with no lockstep release), emitting the existing bundle-shadowed warning - new bundle-discovery-malformed warning code, emitted on warnings (never flips ok / aborts startup) for unresolvable declarations and invalid provider folder names - exported via loader/__init__ and the lfx.extension PEP-562 lazy surface 10 new tests in test_load_lfx_bundles.py; full extension unit suite (449) passes; ruff + format clean; mypy clean on the new module.
|
Important Review skippedAuto incremental reviews are disabled on this repository. Please check the settings in the CodeRabbit UI or the ⚙️ Run configurationConfiguration used: Path: .coderabbit.yaml Review profile: CHILL Plan: Pro Run ID: You can disable this status message by setting the Use the checkbox below for a quick retry:
WalkthroughThis PR adds manifest-less bundle discovery for the ChangesManifest-less Bundle Discovery
Estimated code review effort🎯 4 (Complex) | ⏱️ ~45 minutes Suggested labels
Suggested reviewers
🚥 Pre-merge checks | ✅ 8 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (8 passed)
✏️ Tip: You can configure your own custom pre-merge checks in the settings. ✨ Finishing Touches🧪 Generate unit tests (beta)
Thanks for using CodeRabbit! It's free for OSS, and your support helps us grow. If you like it, consider giving us a shout-out. Comment |
github-actions
Bot
added
enhancement
…ndation-discovery
github-actions
Bot
added
enhancement
The BUNDLE_API.md changelog gate diffs each PR's branch against main, so the entry covering this PR's surface additions (lfx.bundles discovery, the lfx_bundles precedence tier, bundle-discovery-malformed) must live on this branch, not only further up the stack. Text is verbatim from the graduate-partners commit so the stacked merge dedupes trivially.
…ndation-discovery
github-actions
Bot
added
enhancement
github-actions
Bot
added
enhancement
✅ Test Coverage AdvisorNo source changes detected without accompanying tests. Thanks for keeping coverage up! 🎉
|
github-actions
Bot
added
enhancement
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
There was a problem hiding this comment.
Actionable comments posted: 2
🤖 Prompt for all review comments with AI agents
Verify each finding against current code. Fix only still-valid issues, skip the
rest with a brief reason, keep changes minimal, and validate.
Inline comments:
In `@BUNDLE_API.md`:
- Around line 405-421: Update the BUNDLE_API.md wording for manifest-less
lfx.bundles to avoid implying the CLI can validate those directories: change
“exempt from `lfx extension validate`” to something like “not validated / not
intended as input to `lfx extension validate`” and add a one-line explanation
that validate_extension() (the CLI check) requires an extension.json or
[tool.langflow.extension] in pyproject.toml so pointing the validator at a
manifest-less provider will report “manifest-not-found”; reference the loader
module src/lfx/src/lfx/extension/loader/_bundles_root.py and the
validate_extension() CLI behavior in the note.
In `@src/lfx/src/lfx/extension/loader/_bundles_root.py`:
- Around line 246-260: _spec_package_dir currently treats plain module specs as
package roots by returning Path(spec.origin).parent when
submodule_search_locations is absent; change it to reject non-package
entry-point targets by removing the fallback that returns the origin's parent.
Update _spec_package_dir so it only returns a Path when
spec.submodule_search_locations is present (i.e., a package or namespace
package) and otherwise returns None; keep the initial spec is None guard and do
not treat spec.origin as a package root.
🪄 Autofix (Beta)
Fix all unresolved CodeRabbit comments on this PR:
- Push a commit to this branch (recommended)
- Create a new PR with the fixes
ℹ️ Review info
⚙️ Run configuration
Configuration used: Path: .coderabbit.yaml
Review profile: CHILL
Plan: Pro
Run ID: adefeccc-6f5e-4660-b680-b1263385f836
📒 Files selected for processing (8)
BUNDLE_API.mdsrc/lfx/src/lfx/extension/__init__.pysrc/lfx/src/lfx/extension/errors.pysrc/lfx/src/lfx/extension/loader/__init__.pysrc/lfx/src/lfx/extension/loader/_bundles_root.pysrc/lfx/src/lfx/interface/components.pysrc/lfx/tests/unit/extension/loader/test_load_lfx_bundles.pysrc/lfx/tests/unit/extension/test_errors.py
…date wording CodeRabbit review fixes on #13563: - _spec_package_dir no longer falls back to a plain module's parent directory -- a single-file entry-point target now emits the bundle-discovery-malformed warning instead of folder-walking unrelated sibling directories as bundles (+ regression test) - BUNDLE_API.md / loader docstring no longer say manifest-less providers are 'exempt from lfx extension validate'; they are not valid validator input (validate requires a manifest and reports manifest-not-found)
github-actions
Bot
added
enhancement
ogabrielluiz
left a comment
ogabrielluiz
left a comment
There was a problem hiding this comment.
Hey @erichare, re-reviewed after the head moved. The six _bundles_root.py / errors.py notes from my last pass are all addressed: the dotted-find_spec parent import now catches broadly, the same-tier duplicate uses duplicate-lfx-bundles-provider instead of the cross-source bundle-shadowed template, the claimed-bundle skip-before-import avoids overwriting the winner's _lfx_ext.official.<bundle>.* modules, _spec_package_dirs walks every namespace portion with resolved-path dedup, and the plain-module case returns []. The two freeze-gate notes are covered too, the gate now runs on pull_request in extension-migration-checks.yml. Nice cleanup.
Two small things left, both docs:
The errors.py comment says the new lfx.bundles codes are "All warning-only" and "mirror the inline tier's split", but inline-bundle-name-invalid and inline-path-unreadable go on result.errors (so ok=False), while bundles-provider-name-invalid and bundles-root-unreadable go on .warnings (so ok=True). Only duplicate-inline-bundle actually matches. The behavior is fine, but the comment overclaims the mirror. Could you tighten it to say only the duplicate case mirrors, or move the two name/readable codes to .errors to match inline?
Manifest-less lfx.bundles providers still don't show up in lfx extension list. discover_all_extensions only walks installed + seed (both manifest-only) and never calls load_lfx_bundles_extensions, which is startup-only, so a live metapackage-served provider is invisible at list-time and the shadow only resolves at server startup. That's a reasonable design given the sys.modules namespace is assembled at startup, but BUNDLE_API.md's manifest-less block doesn't mention the list-time limitation. Could you add a line so the precedence story isn't read as list-time too?
| "Groq": {"langchain-groq"}, | ||
| "IBM watsonx.ai": {"langchain-ibm"}, | ||
| "Ollama": {"langchain-ollama"}, | ||
| "OpenAI": {"langchain-openai"}, |
There was a problem hiding this comment.
Hey @erichare, I think the preflight and the shim send users to two different packages here. On an engine-only lfx install, a flow with an OpenAI component goes like this: lfx run preflight hits this fallback and says pip install langchain-openai, the user installs it and re-runs, then graph load fails on the import shim with pip install lfx-openai. Two round-trips, and langchain-openai alone never fixes it because the missing piece is the component package, not the SDK.
Could the preflight name the lfx-<provider> / lfx-bundles distribution that's actually missing, since that's the engine-only case it's meant to protect? Or is the shim error meant to be the authoritative one and this is just a best-effort SDK hint?
There was a problem hiding this comment.
I dug into this before changing it, and I think the SDK is actually the right target for this table — flipping it to the bundle would send the wrong package on the path this table serves.
_PROVIDER_PACKAGE_FALLBACKS is consulted only via _detect_providers_from_template, i.e. the unified model / agent_llm / embeddings_model provider dropdowns. At runtime that path is:
LanguageModelComponent.build_model → get_llm → get_model_class("ChatOpenAI") → class_registry._MODEL_CLASS_IMPORTS["ChatOpenAI"] = ("langchain_openai", "ChatOpenAI", None)
so it imports langchain_openai directly and never touches lfx.components.openai — the shim doesn't fire on this path, and pip install langchain-openai makes the flow run in one step. The same table also feeds the wxO deployment artifact (build_langflow_artifact_bytes), whose runner builds models the same way, so naming the bundle there would drag in the whole bundle when only the SDK is needed.
The two-round-trip / shim-firing case you saw sounds like a standalone provider component loaded by module reference — a different path (static code analysis), which also already resolves to the SDK. If you've got that flow handy I'm happy to dig into that path separately, since it's a real wart even if it isn't this table. So I've left the table as-is for now; let me know if you'd still prefer the bundle name and I'll reconcile it with the wxO export expectations.
The three doc points are fixed in f516062: this comment's inline-mirror overclaim in errors.py, the lfx extension list list-time limitation in BUNDLE_API.md, and the lfx-bundles README's "empty skeleton" framing.
… metapackage README Addresses three documentation nits from the PR #13563 re-review: - errors.py: the lfx.bundles diagnostics comment claimed a full mirror of the inline tier's split, but only the duplicate case matches. name-invalid and root-unreadable are warning-only here where their inline counterparts are hard errors (ok=False). Tighten the comment to match actual behavior. - BUNDLE_API.md: note that manifest-less lfx.bundles providers are invisible to `lfx extension list` (discover_all_extensions walks only installed + seed; load_lfx_bundles_extensions runs in the startup component-loading path), so the precedence story isn't read as a list-time property. - lfx-bundles/README.md: the bulk move has landed on this branch (71 providers, populated per-provider extras, generated `all`), so drop the "empty skeleton / available later" framing and document the final state. Co-Authored-By: Claude Opus 4.8 <noreply@anthropic.com>
|
Ran a probe on this branch with engine-only The actionable install guidance only exists on two routes: the legacy import shim, and the Repro: load The graduated providers are fine, for what it's worth. A flow importing Could we catch |
…rver run path A flow loaded through the UI/API on an engine-only `lfx` (no bundles) failed deep in the vertex build with a bare `Error creating class. ModuleNotFoundError(No module named 'langchain_chroma')`. The actionable install guidance only existed on the `lfx run` preflight and the legacy import shim; the server run path (simple_run_flow -> run_graph_internal -> graph.arun()) had neither. Both vertex-build error formatters (lfx's `format_exception_message`, used by graph.arun(), and langflow's, used by the build/chat endpoints) now detect a plain ModuleNotFoundError in the cause chain and attach the same `pip install <package>` guidance the CLI gives, mapping the missing import to its PyPI distribution via the existing requirements resolver. Bundle-shim "components moved to ..." messages are left untouched (their message does not start with "No module named"), so graduated providers stay actionable.
|
Good catch — fixed in 1e2395a. Went with option 1 (catch
The graduated-provider shim messages are deliberately left untouched — their |
|
Verified on 1e2395a, the SDK case is fixed. Loading One thing on the graduated case. The commit says the shim "components moved to ..." messages are left untouched, but It unwraps past [1] to [2], so the hint fires and the message gets regenerated as |
|
Build successful! ✅ |
…ption formatter get_causing_exception walked to the deepest __cause__, so a graduated provider's curated "components moved to ..." ModuleNotFoundError (raised `from` the raw "No module named 'lfx_<x>'") was unwrapped past it. The formatter then regenerated bare "pip install lfx-<x>" guidance, dropping the curated text — including the "or pip install langflow, which bundles it" clause. Both get_causing_exception copies (lfx + langflow) now stop at the first ModuleNotFoundError whose message is not a raw "No module named ..." (the curated shim error), and module_not_found_hint surfaces that curated message verbatim. Plain SDK import errors are unchanged. Adds regression tests reproducing the real 3-level cause chain on both formatters; the old test used a truncated chain that hid this.
|
@ogabrielluiz good catch — done in f7b9cb3, exactly your suggestion.
Also added regression tests with the real 3-level chain on both formatters — the old shim test used a truncated chain (curated error with no raw cause underneath), which is why it was green and hid this. |
|
Build successful! ✅ |
…s detection (#13732) fix(bundles): cover OpenRouter/IBM WatsonX in wxO requirements detection The bundle split removed the langchain provider SDKs from lfx's own dependency tree, which makes flow→requirements detection load-bearing for Watsonx Orchestrate deployments: the runner installs only `lfx` + the packages `generate_requirements_from_flow` emits, so a provider missing from the resolution tables yields an empty requirements set and the deployed flow fails to import its model class. Two unified-selectable providers were unresolved: - OpenRouter (runs on ChatOpenAI) was in no resolution table -> []. - "IBM WatsonX" (catalog casing) only matched the legacy "IBM watsonx.ai" key; langchain-ibm landed only by an embedding-path coincidence. Add both to _PROVIDER_PACKAGE_FALLBACKS (keyed by the exact catalog strings) so every provider in MODEL_PROVIDER_METADATA resolves. Tests: - TestBundleSeparationOrchestrate in test_flow_requirements.py: catalog coverage guard, OpenRouter/IBM regressions, and split-contract guards (bundle dists and langchain imports must stay out of lfx's tree). - End-to-end artifact tests in test_watsonx_orchestrate.py asserting a bundle-provider flow emits the langchain SDK (not lfx-openai/lfx-bundles) and that requirements.txt pins lfx as its first line. Co-authored-by: Claude Opus 4.8 <noreply@anthropic.com>
|
Build successful! ✅ |
…-of-war The datastax bundle's pyproject.toml declares a [tool.ruff.lint.per-file-ignores] table, which makes ruff treat the directory as a standalone config root. It never restated line-length, so ruff fell back to its default of 88 there. As a result `make format_backend` (per-directory config discovery) wrapped lines to 88 while the pre-commit ruff-format hook (run with --config pyproject.toml, forcing the repo-wide 120) unwrapped them back, leaving the two formatters fighting forever. Restate line-length = 120 so both config-resolution paths agree. Use the explicit line-length rather than extend=../../../pyproject.toml to avoid pulling in the root lint ruleset, which the bundle intentionally does not enforce.
|
Build successful! ✅ |
1 similar comment
|
Build successful! ✅ |
|
Build successful! ✅ |
|
Hey @erichare, quick one on the precedence chain. LE-1491 lists langflow.plugins as a component tier below lfx.bundles, but as far as I can tell langflow.plugins is only read in plugin_routes.py for route registrars now. Nothing loads its component entry-points into the palette, so there's no lfx.bundles-vs-plugins collision to resolve, and filter_component_entry_points just keeps a manifest dist's component EP from being called as a route registrar. Is langflow.plugins still meant to be a component source somewhere I'm missing? I don't think it's a problem, just want to make sure the spec tier isn't pointing at something real we skipped. |
No, you're not missing a component source. The LE-1491 tier is describing a precedence relationship (manifest is source-of-truth over a dist's legacy langflow.plugins component EPs), but the losing side of that relationship has no live palette consumer — so there's genuinely no lfx.bundles-vs-plugins collision to resolve at the component level. The only residual handling is the plugin_routes guard, which is doing the conservative thing. |
Both providers were dropped during the bulk long-tail move with no bundle
and no compatibility shim, removing them from the registry,
component_index, migration table and locales. That breaks saved flows
using the NotDiamond Router or VLM Run Transcription nodes (vlmrun ships
in langflow[complete]), contradicting the zero-user-facing-change goal.
Relocate both the same way as the other long-tail providers:
- move implementations into src/bundles/lfx-bundles/src/lfx_bundles/
- restore lfx.components.{notdiamond,vlmrun} marker shims
- re-register both in lfx/components/__init__.py
- add per-provider extras (vlmrun -> vlmrun[all]) and the 'all' aggregate
- add migration_table targets ext:{notdiamond,vlmrun}:<Class>@official
- restore the 32 en.json strings; refresh uv.lock
The preflight is wired into 'lfx run'; 'lfx serve' has its own module-not-found hint at request time and does not call it.
ogabrielluiz
left a comment
ogabrielluiz
left a comment
There was a problem hiding this comment.
Approving on the DX and backward-compat front. Everything I raised is resolved and I verified the fixes by running them on the branch, not just reading the commits:
- Missing-SDK deps now surface
pip install <pkg>guidance on the server/UI run path, not justlfx run(1e2395a). - Graduated-provider errors keep the curated "moved to lfx-" shim message, while plain SDK deps get regenerated guidance (f7b9cb3).
- Docs fixes: README install state, the error-tier mirror claim, the list-time discovery note (f516062).
Stock 1.10/1.9 flows on core components load unchanged engine-only, and the langflow.plugins tier question checked out as a precedence description with no live palette consumer, so nothing real is being skipped there.
Scope note: this was the DX and user-confusion lens plus the discovery/precedence, shim/migration, and preflight areas. I didn't line-by-line audit all 573 files for correctness, so this approval is on that scope.
|
Build successful! ✅ |
3 participants
Bundle Separation Phase A — metapackage split (1.11) · meta PR
This is the single PR for the entire Phase A bundle-separation epic. The original 10-PR stack has been fully collapsed into this branch as each ticket passed review — #13564, #13565, #13566, #13567, and #13568 merged here directly, and #13568 itself carried #13573, #13576, #13577, #13578, #13579, and #13580. Only this PR remains to merge into
release-1.11.0.Includes Oracle (#13502): the
lfx-oraclebundle now ships in this branch as a sixth graduated package following the same conventions (manifest-shipping, bounded ranges,.dev0lfx floor, migration-table entries, pilot-upgrade test). An earlier plan to land Oracle as a separate follow-up after this PR was changed — it is folded in here, and the lockfile, Docker images, and workspace members reflect that.What it does:
lfxbecomes engine-only. The ~45-provider long tail moves into one manifest-lesslfx-bundlesmetapackage discovered via a newlfx.bundlesentry-point group, and five partner/flagship providers graduate to standalone manifest-shipping packages:lfx-openai,lfx-anthropic,lfx-amazon,lfx-datastax,lfx-cohere. Zero user-facing change:ext:<provider>:<Class>@officialcomponent ids are unchanged, in-tree shims keepfrom lfx.components.<provider> import Xworking, and existing flows load and run identically.~328 files changed (+39k / −59k with rename detection; most of the delta is providers relocating from
src/lfx/src/lfx/components/intosrc/bundles/).Part 1 — manifest-less
lfx.bundlesdiscovery + precedence tier (foundation)load_lfx_bundles_extensions(src/lfx/src/lfx/extension/loader/_bundles_root.py) iterates thelfx.bundlessetuptools entry-point group, resolves each declared package viafind_spec, folder-walks its immediate subdirectories, and registers each as a bundle at the@officialslot with noextension.json(the langchain-community model). Reuses_load_bundle_directory, mirrorsdiscover_inline_bundles.installed > seed > lfx_bundles > dev > inline. Manifest always wins: a manifest-shippinglfx-<provider>shadows the same-named metapackage provider with a typedbundle-shadowedwarning, and the shadowed provider is never imported — this is what lets a provider graduate with no lockstep release.bundle-discovery-malformederror code, emitted onwarnings(never flipsok/ aborts startup), with per-mode codes for unresolvable declarations, plain-module targets, namespace-package portions, and invalid provider folder names. Manifest-less bundles get a typed reload refusal instead of a silent no-op.Part 2 —
lfx-bundlesmetapackage skeleton (#13564) + 45-provider move (#13568)src/bundles/lfx-bundles/: singlepyproject.tomldeclaring thelfx.bundlesentry-point, a pre-release-safelfx>=1.11.0.dev0,<2.0.0floor, a generatedallextra, and a README documenting the model + install stories. Hyphen dir name sorelease.yml'ssrc/bundles/*/pyproject.tomlglob builds it with zero workflow change.src/lfx/src/lfx/components/intosrc/bundles/lfx-bundles/src/lfx_bundles/, each discovered manifest-less at@official. In-tree marker shims preserve the old import paths; the migration table andcomponent_index.jsonare updated append-only/surgically with sha256 re-verified. This set includesnotdiamondandvlmrun, which an earlier revision of this branch had dropped without a bundle or shim; both are now relocated like the rest (bundle package + shim + migration-table targets + restored locale strings) so saved flows using the NotDiamond Router or VLM Run Transcription nodes keep loading unchanged.scripts/ci/update_bundle_versions.py(+ tests) — the nightly bundle-rename track it served was retired by the canonical pre-release cutover (src/bundles/NIGHTLY.md).Part 3 — engine-only
lfx+lfx[bundles]extra (#13565)pip install lfx→ engine only;pip install "lfx[bundles]"→ engine + thelfx-bundleslong tail. Intentionally nolfx[all].lfx-bundlesappears only underextra == "bundles", never in coreRequires-Dist.Part 4 — partner graduations (#13573)
lfx-openai,lfx-anthropic,lfx-amazon,lfx-datastax,lfx-cohere: each shipsextension.json(lfx.compat: ["1"]), alangflow.extensionsentry-point, anlfxpin, and is a workspace member pinned by langflow. Partner set is disjoint fromlfx-bundles(verified).lfx.base.datastax→lfx_datastax.base, 74 backend tests move with it, 10 repo consumers rewritten (incl. thevector_store_ragstarter project).validate_extensionnow accepts classes whose base is a derived Component base (LCVectorStoreComponentetc.) that inherit class-leveloutputs.lfx-*package names (the engine split plus the graduated bundles — includinglfx-bundlesandlfx-oracle) and, for the Oracle bundle, theoracledbandlangchain-oracledbthird-party packages; no other third-party packages are added or removed.Part 5 — CI: cross-bundle matrix (#13566) + components freeze gate (#13567)
.github/workflows/cross-bundle-test.yml: per (bundle × Python 3.10/3.13) — installs in-repolfx+ the bundle, imports the entry-point package, assertslfx.bundlesdiscovery is error-free, runslfx extension validatefor manifest bundles, and runs the bundle's owntests/. Discovers bundles via the same globrelease.ymluses, so each new extraction self-registers. Triggers:pull_request,workflow_dispatch, weeklyschedule,workflow_call.scripts/ci/check_components_frozen.py+ baselinefrozen_component_dirs.txt: no new top-level provider directory may be added tosrc/lfx/src/lfx/components/— new providers go tolfx-bundlesor a graduated package. Removals are allowed (shim cleanup never trips it). Wired intoextension-migration-checks.ymlso it actually runs on PRs (re-homed fromlint-py.ymlduring review).Part 6 — hardening + review follow-ups (#13576–#13580 and branch commits)
lfx-*packages (chore(bundles): bounded version ranges for curated lfx-* packages #13576); shim-contract lock tests + breakage audit (test(bundles): lock the in-tree shim contract + breakage audit #13577); bundle docs — install shapes, override rule,bundle_api_version(docs(bundles): install shapes, override rule, bundle_api_version #13578); stack-review fixes — symlink containment, idempotency, docs accuracy (fix(bundles): stack-review fixes — symlink containment, idempotency, docs accuracy #13579); tombstone for the broken legacyZepChatMemorybuild method (fix(bundles): tombstone the broken legacy ZepChatMemory build method #13580).loading.py/tableAutoCellRenderto base-branch state (out-of-scope changes reverted); frontend test updates.a05d937): the static→bundle move made two pieces of state order-dependent. (1)flatten_components_with_aliasesnow registers every component's identity aliases (registry key / ext class name /name) before any component'sdisplay_namealias, so a standalone component always wins its own type key over a Composio wrapper that merely shares its display label. A singlesetdefaultpass previously let this resolve by registry load order, soupdate_starter_projectswrote the wrongmetadata.modulefor the AgentQL node inNews Aggregator.json/Price Deal Finder.json(redTest Starter Templates/test-starter-projects); both are corrected back tolfx.components.agentql.agentql_api.AgentQL, and an audit confirmedAgentQLandSerpAPIare the only such display-name collisions and both now resolve correctly. (2)SIDEBAR_BUNDLESnameslugs for Notion and FAISS were still capitalized and so never matched the lowercase bundle category names — both bundles were silently dropped from the sidebar (redintegration-side-barPlaywright e2e); lowercased tonotion/faiss(display labels and icons unchanged). Order-independence regression tests added.Test plan
@officialregistration, malformed-declaration / plain-module / namespace-portion / invalid-name warnings, manifest-shadows-manifestless, typed reload refusallfx_bundlesregisters its providers, disjoint from the partner setimport lfx.componentslfx extension validatepasses for all manifest bundles;component_index.jsonsurgically updated, sha256 verifieduv syncresolves; wheels build;lfxMETADATA carrieslfx-bundlesonly underextra == "bundles"AgentQL&SerpAPIresolve to the standalone component regardless of registry load order; 236/236 starter-projects template tests pass; bundles sidebar renders Notion + FAISSCollapsed stack (all merged into this branch)
lfx-bundlesmetapackage skeletonlfx[bundles]extra, engine-onlylfxlfx/componentsfreeze gatelfx-*bundle_api_versionZepChatMemorybuildSummary by CodeRabbit