Add HuggingFace Inference and code optimization #20

10234567Z · 2026-01-13T05:52:26Z

Removed exposed/fallback API Keys from nanda_adapter/core/agent_bridge.py and nanda_adapter/core/mcp_utils.py
Abstract the LLM Provider to easily be able to add support for multiple types of agents in future.
Adds HuggingFace Inference Provider Integration so that Hugging Face API Keys work too.
Fixed requirement.txt i.e dependency listing file by removing duplicates and unused dependencies.

Summary: The code is now secure, optimized, abstract for future use and now even supports HuggingFace Inference for any person wanting to run low cost inference models.

Copilot

Pull request overview

This PR adds multi-provider LLM support to the NANDA framework, removes exposed API keys from code, and optimizes dependencies. The changes introduce an abstraction layer that allows using either Anthropic Claude or Hugging Face Inference API, improving flexibility and security.

Changes:

Created new llm_providers.py module with provider abstraction supporting Anthropic and HuggingFace
Removed hardcoded fallback API keys from agent_bridge.py and mcp_utils.py
Updated requirements.txt to remove duplicates and add huggingface_hub dependency
Modified nanda.py to initialize and configure LLM providers with new optional parameters
Refactored mcp_utils.py and agent_bridge.py to use provider abstraction instead of direct Anthropic SDK calls
Updated documentation in README.md with HuggingFace usage examples and configuration details

Reviewed changes

Copilot reviewed 7 out of 9 changed files in this pull request and generated 20 comments.

Show a summary per file

File	Description
requirements.txt	Removed duplicate anthropic and pymongo entries, added huggingface_hub
nanda_adapter/core/llm_providers.py	New provider abstraction with AnthropicProvider and HuggingFaceProvider classes
nanda_adapter/core/nanda.py	Added LLM provider initialization with new parameters for provider selection
nanda_adapter/core/mcp_utils.py	Refactored to use provider abstraction, handles both dict and object response formats
nanda_adapter/core/agent_bridge.py	Removed hardcoded API keys, updated to use provider abstraction, fixed typo
nanda_adapter/core/init.py	Exported new LLM provider classes and functions
nanda_adapter/init.py	Exported new LLM provider classes and functions
README.md	Added HuggingFace documentation and updated environment variable documentation
.gitignore	Added wildcard pattern for pycache directories

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-01-13T06:02:47Z

nanda_adapter/core/llm_providers.py

+                    content.append({
+                        "type": "tool_use",
+                        "id": tool_call.id,
+                        "name": tool_call.function.name,
+                        "input": json.loads(tool_call.function.arguments) if isinstance(tool_call.function.arguments, str) else tool_call.function.arguments


The error handling for JSON parsing of tool call arguments could be improved. If json.loads fails but the arguments are already a dict, the code should handle this more robustly. Consider adding explicit error handling for the JSON parsing operation to prevent potential runtime errors when tool_call.function.arguments is neither a valid JSON string nor a dict.

Suggested change

content.append({

"type": "tool_use",

"id": tool_call.id,

"name": tool_call.function.name,

"input": json.loads(tool_call.function.arguments) if isinstance(tool_call.function.arguments, str) else tool_call.function.arguments

raw_arguments = tool_call.function.arguments

# Robustly handle tool call arguments which may be a JSON string, dict, or other types

if isinstance(raw_arguments, dict):

parsed_arguments = raw_arguments

elif isinstance(raw_arguments, str):

try:

parsed_arguments = json.loads(raw_arguments)

except Exception:

# Fall back to the raw string if JSON parsing fails

parsed_arguments = raw_arguments

else:

# For unexpected types, pass through as-is

parsed_arguments = raw_arguments

content.append({

"type": "tool_use",

"id": tool_call.id,

"name": tool_call.function.name,

"input": parsed_arguments

Copilot · 2026-01-13T06:02:47Z

nanda_adapter/core/agent_bridge.py

+        if response_text:
+            # Log the LLM response
+            log_message(conversation_id, current_path, f"{provider.name} {agent_id}", response_text)



Inconsistent error handling for missing response_text. In the updated code at line 178, the function checks 'if response_text:' before logging, but still returns response_text regardless. If response_text is None (which can happen from provider.complete() failures), the function should return None explicitly or handle the None case more clearly for the caller.

Suggested change

if response_text:

# Log the LLM response

log_message(conversation_id, current_path, f"{provider.name} {agent_id}", response_text)

if response_text is None:

# Handle missing response_text from provider as a failure case

print(f"Agent {agent_id}: {provider.name} returned no response_text", flush=True)

return None

# Log the LLM response

log_message(conversation_id, current_path, f"{provider.name} {agent_id}", response_text)

Copilot · 2026-01-13T06:02:47Z

nanda_adapter/core/mcp_utils.py

+            import traceback
+            traceback.print_exc()


The traceback module is used at line 201 but is not imported at the top of the file. This will cause a NameError at runtime when an exception occurs. Add 'import traceback' to the imports section at the top of the file.

Copilot · 2026-01-13T06:02:48Z

nanda_adapter/core/mcp_utils.py

+                content = response.get("content", [])
+
+                # Handle different content formats (Anthropic raw vs unified)
+                if hasattr(content, '__iter__') and not isinstance(content, (str, dict)):


Potential issue with iteration check. The condition 'hasattr(content, 'iter')' will be True for strings since strings are iterable in Python. The additional check 'not isinstance(content, (str, dict))' prevents this, but the logic could be clearer. Consider simplifying this to check if content is a list directly: 'isinstance(content, list)' would be more explicit and readable.

Suggested change

if hasattr(content, '__iter__') and not isinstance(content, (str, dict)):

if isinstance(content, list):

Copilot · 2026-01-13T06:02:48Z

nanda_adapter/core/mcp_utils.py

+            if hasattr(content, '__iter__') and not isinstance(content, (str, dict)):
+                blocks = content
+            else:
+                blocks = [content] if content else []


Duplicate code pattern for content iteration. The same logic for checking if content is iterable and converting it to blocks appears at lines 110-113 and again at lines 183-186. This duplicated logic should be extracted into a helper function to improve maintainability.

Copilot · 2026-01-13T06:02:51Z

nanda_adapter/core/agent_bridge.py

+    from .llm_providers import get_provider, init_provider
+    from .mcp_utils import MCPClient
+except ImportError:
+    from llm_providers import get_provider, init_provider


Import of 'init_provider' is not used.

Suggested change

from .llm_providers import get_provider, init_provider

from .mcp_utils import MCPClient

except ImportError:

from llm_providers import get_provider, init_provider

from .llm_providers import get_provider

from .mcp_utils import MCPClient

except ImportError:

from llm_providers import get_provider

Copilot · 2026-01-13T06:02:51Z

nanda_adapter/core/agent_bridge.py

+    from .llm_providers import get_provider, init_provider
+    from .mcp_utils import MCPClient
+except ImportError:
+    from llm_providers import get_provider, init_provider


Import of 'init_provider' is not used.

Suggested change

from .llm_providers import get_provider, init_provider

from .mcp_utils import MCPClient

except ImportError:

from llm_providers import get_provider, init_provider

from .llm_providers import get_provider

from .mcp_utils import MCPClient

except ImportError:

from llm_providers import get_provider

Copilot · 2026-01-13T06:02:51Z

nanda_adapter/core/nanda.py

+    from .llm_providers import init_provider, get_provider
 except ImportError:
    # If running from parent directory, add current directory to path
    current_dir = os.path.dirname(os.path.abspath(__file__))
    sys.path.insert(0, current_dir)
    from agent_bridge import *
    import run_ui_agent_https
+    from llm_providers import init_provider, get_provider


Import of 'get_provider' is not used.

Suggested change

from .llm_providers import init_provider, get_provider

except ImportError:

# If running from parent directory, add current directory to path

current_dir = os.path.dirname(os.path.abspath(__file__))

sys.path.insert(0, current_dir)

from agent_bridge import *

import run_ui_agent_https

from llm_providers import init_provider, get_provider

from .llm_providers import init_provider

except ImportError:

# If running from parent directory, add current directory to path

current_dir = os.path.dirname(os.path.abspath(__file__))

sys.path.insert(0, current_dir)

from agent_bridge import *

import run_ui_agent_https

from llm_providers import init_provider

nanda_adapter/core/nanda.py

Copilot · 2026-01-13T06:02:52Z

nanda_adapter/core/agent_bridge.py

-SMITHERY_API_KEY = os.getenv("SMITHERY_API_KEY") or "bfcb8cec-9d56-4957-8156-bced0bfca532"
+SMITHERY_API_KEY = os.getenv("SMITHERY_API_KEY")
+if not SMITHERY_API_KEY:
+    print("WARNING: SMITHERY_API_KEY not set - Smithery MCP servers will not work")


Print statement may execute during import.

Co-authored-by: Copilot <[email protected]>

10234567Z added 5 commits January 12, 2026 19:13

Remove fallback/exposed API Keys

4dff4ce

Remove double record of anthropic library

548b49c

Remove unused pymongo library

a52d9f4

Add support for Huggingface Inference

54a320c

Update README to document HF

4dea3dc

Copilot AI review requested due to automatic review settings January 13, 2026 05:52

Copilot started reviewing on behalf of 10234567Z January 13, 2026 05:52 View session

Copilot AI reviewed Jan 13, 2026

View reviewed changes

Update nanda_adapter/core/nanda.py

8dca7cd

Co-authored-by: Copilot <[email protected]>

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Add HuggingFace Inference and code optimization #20

Add HuggingFace Inference and code optimization #20

Uh oh!

10234567Z commented Jan 13, 2026

Uh oh!

Copilot AI left a comment

Uh oh!

Copilot AI Jan 13, 2026

Uh oh!

Copilot AI Jan 13, 2026

Uh oh!

Copilot AI Jan 13, 2026

Uh oh!

Copilot AI Jan 13, 2026

Uh oh!

Copilot AI Jan 13, 2026

Uh oh!

Copilot AI Jan 13, 2026

Uh oh!

Copilot AI Jan 13, 2026

Uh oh!

Copilot AI Jan 13, 2026

Uh oh!

Uh oh!

Copilot AI Jan 13, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant

-                    content.append({
-                        "type": "tool_use",
-                        "id": tool_call.id,
-                        "name": tool_call.function.name,
-                        "input": json.loads(tool_call.function.arguments) if isinstance(tool_call.function.arguments, str) else tool_call.function.arguments
+                    raw_arguments = tool_call.function.arguments
+                    # Robustly handle tool call arguments which may be a JSON string, dict, or other types
+                    if isinstance(raw_arguments, dict):
+                        parsed_arguments = raw_arguments
+                    elif isinstance(raw_arguments, str):
+                        try:
+                            parsed_arguments = json.loads(raw_arguments)
+                        except Exception:
+                            # Fall back to the raw string if JSON parsing fails
+                            parsed_arguments = raw_arguments
+                    else:
+                        # For unexpected types, pass through as-is
+                        parsed_arguments = raw_arguments
+                    content.append({
+                        "type": "tool_use",
+                        "id": tool_call.id,
+                        "name": tool_call.function.name,
+                        "input": parsed_arguments

	if hasattr(content, '__iter__') and not isinstance(content, (str, dict)):
	if isinstance(content, list):

Add HuggingFace Inference and code optimization #20

Are you sure you want to change the base?

Add HuggingFace Inference and code optimization #20

Uh oh!

Conversation

10234567Z commented Jan 13, 2026

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Jan 13, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 13, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 13, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 13, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 13, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 13, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 13, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 13, 2026

Choose a reason for hiding this comment

Uh oh!

Uh oh!

Copilot AI Jan 13, 2026

Choose a reason for hiding this comment

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

1 participant