fix: resolve embedding-model fallbacks from deployment env, never guess (backport of #2191) - #2192
Merged
edwinjosechittilappilly merged 2 commits intoJul 31, 2026
Conversation
_default_embedding_model() previously hardcoded a per-provider guess (e.g. "text-embedding-3-small" for openai) used whenever a provider removal left another provider as the embedding fallback. A provider name doesn't reliably say which models are actually deployed — it's often an internal OpenAI/watsonx-compatible gateway with a curated model subset. The hardcoded guess silently selected a model the gateway didn't serve, breaking every ingestion with no clear signal why (incident: gateway only served "text-embedding-3-large"; 25/25 files failed at the Langflow embedding step). Add get_declared_default_embedding_model(), which resolves the deployment's own declared default from EMBEDDING_MODEL / EMBEDDING_PROVIDER (Helm values / operator ConfigMap) instead of a hardcoded string, and returns "" when the deployment hasn't declared one for that provider. Read directly rather than through ConfigManager's env-override path, which is skipped once config.edited is True — a protection meant for a user's own prior choices, not a reason to block resolving an unmade choice. Wire this into every "guess an embedding model" call site, not just the settings-save path: search_service.py, utils/embeddings.py, and langflow_file_service.py all previously fell back to the same hardcoded constant when no model was configured; they now fall back to the deployment-declared default (or none) instead.
github-actions
Bot
added
backend
Contributor
WalkthroughEmbedding model fallback selection now uses deployment-declared provider and model environment variables. Explicit and configured models remain higher priority. Missing or mismatched declarations return an empty model. ChangesEmbedding default resolution
Estimated code review effort: 3 (Moderate) | ~20 minutes Possibly related PRs
Suggested reviewers: 🚥 Pre-merge checks | ✅ 4 | ❌ 1❌ Failed checks (1 warning)
✅ Passed checks (4 passed)
✨ Finishing Touches 💡 1📝 Generate docstrings 💡
🧪 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
bug
![coderabbitai[bot]](https://avatars.githubusercontent.com/in/347564?s=60&v=4){kind=small}
Contributor
There was a problem hiding this comment.
Actionable comments posted: 1
🤖 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 `@src/config/embedding_constants.py`:
- Around line 3-33: Move the EMBEDDING_PROVIDER and EMBEDDING_MODEL environment
reads out of get_declared_default_embedding_model in embedding_constants.py and
into a dedicated declared-default accessor in config/settings.py. Update
get_declared_default_embedding_model to call that accessor while preserving its
provider matching and empty-result behavior, including fallback resolution after
configuration edits.
🪄 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 Plus
Run ID: b566ae4e-3da2-4973-9502-adfe5228a6f3
📒 Files selected for processing (6)
src/api/settings/helpers.pysrc/config/embedding_constants.pysrc/services/langflow_file_service.pysrc/services/search_service.pysrc/utils/embeddings.pytests/unit/test_settings_provider_removal_defaults.py
Comment on lines
+3
to
+33
| import os | ||
|
|
||
| OPENAI_DEFAULT_EMBEDDING_MODEL = "text-embedding-3-small" | ||
| OPENAI_EMBEDDING_MODEL_PREFIX = "text-embedding" No newline at end of file | ||
| OPENAI_EMBEDDING_MODEL_PREFIX = "text-embedding" | ||
|
|
||
|
|
||
| def get_declared_default_embedding_model(provider: str) -> str: | ||
| """Return the deployment-declared default embedding model for `provider`. | ||
|
|
||
| Sourced directly from the EMBEDDING_MODEL / EMBEDDING_PROVIDER env vars | ||
| the deployment is expected to set (Helm values / operator ConfigMap) to | ||
| describe what this specific environment's embedding backend actually | ||
| serves. Read directly rather than through ConfigManager's env-override | ||
| path, which is skipped once the config has been manually edited — a | ||
| protection against clobbering a user's own choices that shouldn't also | ||
| block resolving a fallback when the user hasn't made one. | ||
|
|
||
| Returns "" when the deployment hasn't declared a default for this | ||
| provider. Callers MUST NOT substitute a hardcoded guess in that case — | ||
| "openai" (and any other provider name) can mean anything from the real | ||
| public API to an internal gateway with a curated model subset, so no | ||
| single hardcoded model name is safe across deployments. A hardcoded | ||
| guess here previously selected "text-embedding-3-small" in an | ||
| environment whose gateway only served "text-embedding-3-large", | ||
| silently failing every ingestion. | ||
| """ | ||
| declared_provider = os.environ.get("EMBEDDING_PROVIDER", "") | ||
| declared_model = os.environ.get("EMBEDDING_MODEL", "") | ||
| if declared_provider == provider and declared_model: | ||
| return declared_model | ||
| return "" |
Contributor
There was a problem hiding this comment.
📐 Maintainability & Code Quality | 🟠 Major | ⚡ Quick win
Move environment reads to config/settings.py.
This module imports os and reads EMBEDDING_PROVIDER and EMBEDDING_MODEL directly. This violates the configuration-source contract. Add a dedicated declared-default accessor in config/settings.py, then call that accessor here. Preserve the required post-edit fallback behavior in that accessor.
As per path instructions, “Config values must come from config/settings.py (the only place os.environ is read); never access os.environ elsewhere in the codebase.”
🤖 Prompt for 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.
In `@src/config/embedding_constants.py` around lines 3 - 33, Move the
EMBEDDING_PROVIDER and EMBEDDING_MODEL environment reads out of
get_declared_default_embedding_model in embedding_constants.py and into a
dedicated declared-default accessor in config/settings.py. Update
get_declared_default_embedding_model to call that accessor while preserving its
provider matching and empty-result behavior, including fallback resolution after
configuration edits.
Source: Path instructions
github-actions
Bot
deleted the
backport/2191-embedding-model-fallback-not-validated-main
branch
July 31, 2026 20:01
This was referenced Aug 3, 2026
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
2 participants
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.
Backport of #2191 to `main`. Clean cherry-pick, no conflicts.
Summary
Test plan
Summary by CodeRabbit
Bug Fixes
Tests