feat: nvidia_nim encoder

[Joshua-Briggs]

PR Type

• Enhancement
• Tests
• Documentation

---

Description

• Added Jina, NVIDIA NIM, Voyage encoder classes.

• Integrated new encoder types in AutoEncoder.

• Updated defaults, schema and tests accordingly.

• Provided new notebooks for encoder usage.

---

Changes walkthrough 📝

	Relevant files
Enhancement	6 files __init__.py Integrate new Jina, Nim, Voyage encoder imports                    +12/-0    jina.py Add new JinaEncoder implementation class                                  +34/-0    nvidia_nim.py Add new NimEncoder for Nvidia NIM service                                +85/-0    voyage.py Add new VoyageEncoder class for Voyage models                        +85/-0    schema.py Extend EncoderType enum with new encoder types                      +10/-0    defaults.py Add default configurations for new encoders                            +12/-0
Bug fix	1 files litellm.py Refine name splitting for encoder instantiation                    +1/-1
Tests	1 files test_lite_encoders.py Add test matrix entries for new encoder types                        +36/-2
Documentation	4 files jina-encoder.ipynb Document Jina encoder usage with notebook                                +417/-0  mistral-encoder.ipynb Update mistral version from 1.1.6 to 0.1.8                              +2/-2      nvidia_nim-encoder.ipynb Include notebook for Nvidia NIM encoder demonstration        +407/-0  voyage-encoder.ipynb Add notebook documenting Voyage encoder usage                        +393/-0

---

Need help?

Type /help how to ... in the comments thread for any questions about PR-Agent usage.

Check out the documentation for more information.

[github-actions]

PR Reviewer Guide 🔍

Here are some key observations to aid the review process:

⏱️ Estimated effort to review: 3 🔵🔵🔵⚪⚪

🧪 PR contains tests

🔒 No security concerns identified

⚡ Recommended focus areas for review Docstring Inconsistency The docstring for the initializer refers to a parameter named "jina_api_key" while the actual parameter is "api_key". Consider aligning the documentation with the code for clarity. type: str = "jina" def __init__( self, name: str | None = None, api_key: str | None = None, score_threshold: float = 0.4, ): """Initialize the JinaEncoder. :param name: The name of the embedding model to use such as "jina-embeddings-v3". :param jina_api_key: The Jina API key. :type jina_api_key: str """ if name is None: name = f"jina_ai/{EncoderDefault.JINA.value['embedding_model']}" elif not name.startswith("jina_ai/"): name = f"jina_ai/{name}" super().__init__( name=name, score_threshold=score_threshold, api_key=api_key, )

[github-actions]

PR Code Suggestions ✨

Explore these optional code suggestions:

Category	Suggestion	Impact
Possible issue	Correct dimension description Update the dimension description to match the actual output from rl.index.dimensions. docs/encoders/nvidia_nim-encoder.ipynb [259-265] -"We do have 256-dimensional vectors. Now let's test them:" +"We do have 1024-dimensional vectors. Now let's test them:" Suggestion importance[1-10]: 7 __ Why: The suggestion corrects a factual mismatch in the notebook's narrative by updating the described vector dimensionality from 256 to 1024, which aligns with the actual output. This improves clarity and accuracy in documentation.	Medium
	Align return value description Revise the explanation to accurately reflect that an unmatched query returns a RouteChoice with a None name. docs/encoders/voyage-encoder.ipynb [358-363] -"In this case, we return `None` because no matches were identified. We always recommend optimizing your `RouteLayer` for optimal performance, you can see how in [this notebook](https://github.com/aurelio-labs/semantic-router/blob/main/docs/06-threshold-optimization.ipynb)." +"In this case, we return a RouteChoice with no match (with `name` set to None) because no valid route was identified. We always recommend optimizing your `RouteLayer` for optimal performance, as shown in [this notebook](https://github.com/aurelio-labs/semantic-router/blob/main/docs/06-threshold-optimization.ipynb)." Suggestion importance[1-10]: 4 __ Why: The suggestion refines the explanatory text to indicate that an unmatched query yields a RouteChoice with a None name rather than a raw None value. While this improves precision in the narrative, its impact is limited to documentation clarity.	Low
General	Fix docstring parameter naming Update the docstring parameter name to match the actual constructor parameter (api_key). semantic_router/encoders/jina.py [13-24] def __init__( self, name: str | None = None, api_key: str | None = None, score_threshold: float = 0.4, ): """Initialize the JinaEncoder. :param name: The name of the embedding model to use such as "jina-embeddings-v3". - :param jina_api_key: The Jina API key. - :type jina_api_key: str + :param api_key: The Jina API key. + :type api_key: str """ Suggestion importance[1-10]: 3 __ Why: This change corrects the incorrect parameter name in the docstring, aligning it with the constructor signature. Although it only affects documentation, it improves clarity.	Low
	Update NimEncoder docstring Change the docstring parameter (nim_api_key) to use the actual parameter name (api_key). semantic_router/encoders/nvidia_nim.py [15-26] def __init__( self, name: str | None = None, api_key: str | None = None, score_threshold: float = 0.4, ): """Initialize the NimEncoder. :param name: The name of the embedding model to use such as "nv-embedqa-e5-v5". - :param nim_api_key: The Nim API key. - :type nim_api_key: str + :param api_key: The Nim API key. + :type api_key: str """ Suggestion importance[1-10]: 3 __ Why: The suggested update replaces the misleading parameter name in the NimEncoder docstring with the correct one, enhancing the documentation's accuracy with minimal impact.	Low
	Correct VoyageEncoder docstring Update the docstring parameter name (voyage_api_key) to match the actual parameter (api_key). semantic_router/encoders/voyage.py [15-26] def __init__( self, name: str | None = None, api_key: str | None = None, score_threshold: float = 0.4, ): """Initialize the VoyageEncoder. :param name: The name of the embedding model to use such as "voyage-embed". - :param voyage_api_key: The Voyage API key. - :type voyage_api_key: str + :param api_key: The Voyage API key. + :type api_key: str """ Suggestion importance[1-10]: 3 __ Why: By updating the docstring parameter from "voyage_api_key" to "api_key," the change improves consistency with the function signature, representing a minor yet useful documentation fix.	Low

[codecov]

Codecov Report

Attention: Patch coverage is 58.53659% with 17 lines in your changes missing coverage. Please review.

Project coverage is 74.25%. Comparing base (900a0c9) to head (47797df).
Report is 7 commits behind head on main.

Files with missing lines	Patch %	Lines
semantic_router/encoders/nvidia_nim.py	57.14%	15 Missing ⚠️
semantic_router/encoders/__init__.py	33.33%	2 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##             main     #582   +/-   ##
=======================================
  Coverage   74.25%   74.25%           
=======================================
  Files          48       48           
  Lines        4373     4374    +1     
=======================================
+ Hits         3247     3248    +1     
  Misses       1126     1126           

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.
• 📦 JS Bundle Analysis: Save yourself from yourself by tracking and limiting bundle sizes in JS merges.

[cursor]

Bug: Incorrect Default Model for Embeddings

The default embedding_model for NVIDIA_NIM is incorrectly set to "meta/llama3-70b-instruct". This model is a language model, not an embedding model, which causes failures when generating embeddings with default settings. The default should be an actual embedding model, such as "nvidia/nv-embedqa-e5-v5", as indicated in documentation and usage examples.

semantic_router/utils/defaults.py#L31-L39

semantic-router/semantic_router/utils/defaults.py

Lines 31 to 39 in a4b7ea7

	}
	NVIDIA_NIM = {
	"embedding_model": os.getenv(
	"NVIDIA_NIM_MODEL_NAME", "meta/llama3-70b-instruct"
	),
	"language_model": os.getenv(
	"NVIDIA_NIM_CHAT_MODEL_NAME", "meta/llama3-70b-instruct"
	),
	}

Fix in Cursor

---

Was this report helpful? Give feedback by reacting with 👍 or 👎