fix multi turn tool calling for anthropic when arguments to the tool call is an empty object #1104
Conversation
…call is an empty object
|
Important PR Review SkippedPR review skipped as per the configuration setting. Run a manually review by commenting /matter review 💡Tips to use Matter AICommand List
|
|
Summary By MatterAI
🔄 What Changed
🔍 Impact of the Change
📁 Total Files Changed
🧪 Test Added
🔒 Security Vulnerabilities
Type of Change
Checklist
Sequence DiagramsequenceDiagram
participant Client
participant AnthropicChatComplete
participant ToolCall
Client->>AnthropicChatComplete: Invoke chat completion
AnthropicChatComplete->>ToolCall: Process tool calls
Note over ToolCall: Check tool function arguments length
alt Arguments exist
ToolCall->>ToolCall: JSON.parse(arguments)
else No arguments
ToolCall->>ToolCall: Return empty object {}
end
AnthropicChatComplete-->>Client: Return processed tool calls
|
narengogi
changed the title
narengogi
requested review from
sabbyasachiRPortkey and
siddharthsambharia-portkey
|
Important PR Review SkippedPR review skipped as per the configuration setting. Run a manually review by commenting /matter review 💡Tips to use Matter AICommand List
|
|
Summary By MatterAI
🔄 What Changed
🔍 Impact of the Change
📁 Total Files Changed
🧪 Test Added
🔒 Security Vulnerabilities
Type of Change
Checklist
Sequence DiagramsequenceDiagram
participant Anthropic as A
participant ToolCall as TC
participant ChatComplete as CC
A->>TC: Prepare Tool Call
TC-->>CC: Pass Tool Call Arguments
CC->>TC: Parse Arguments
Note over CC: Handle Empty Object Case
TC->>A: Return Processed Tool Call
|
|
DescriptionSummary By MatterAI
🔄 What ChangedThis pull request addresses a bug in handling multi-turn tool calls for Anthropic models, specifically when the arguments to a tool call are an empty object. The 🔍 Impact of the ChangeThis change fixes a critical bug that prevented successful multi-turn tool calling with Anthropic when tool arguments were empty, improving the robustness and reliability of the integration. It ensures that tool calls with empty argument objects are correctly processed, preventing parsing errors and allowing the conversation flow to continue as expected. The 📁 Total Files Changed1 file changed. 🧪 Test AddedNo explicit unit or integration tests were mentioned in the PR description. However, it is assumed that manual testing was performed to verify the fix for the specific scenario of empty tool call arguments. 🔒Security VulnerabilitiesNo new security vulnerabilities were detected in the changes. The fix improves robustness by handling a specific input case more gracefully. MotivationThis change is needed to fix a bug where multi-turn tool calling for Anthropic fails when arguments to the tool call is an empty object, as reported in #1103. Type of Change
How Has This Been Tested?
Screenshots (if applicable)N/A Checklist
Related IssuestotalScore: 8 Tip Quality Recommendations
Sequence DiagramsequenceDiagram
participant User as User/Client
participant Gateway as Portkey Gateway
participant AnthropicAPI as Anthropic API
User->>Gateway: Send Chat Completion Request (with tool definitions)
Gateway->>AnthropicAPI: Forward Request (with updated tool definitions including $defs)
AnthropicAPI-->>Gateway: Respond with Assistant Message (potentially with tool_use)
Gateway->>Gateway: Call transformAssistantMessage(msg: Message)
activate Gateway
Gateway->>Gateway: Iterate msg.tool_calls
loop For each toolCall
Gateway->>Gateway: Access toolCall.function.arguments
alt arguments is empty string or null/undefined
Gateway->>Gateway: Set input = {}
else arguments is non-empty string
Gateway->>Gateway: Call JSON.parse(toolCall.function.arguments)
Note right of Gateway: Handles potential malformed JSON if not empty
end
Gateway->>Gateway: Construct AnthropicMessage with tool_use and parsed input
end
deactivate Gateway
Gateway-->>User: Return Transformed Chat Completion Response
|
|
Important PR Review SkippedPR review skipped as per the configuration setting. Run a manually review by commenting /matter review 💡Tips to use Matter AICommand List
|
This PR fixes
$defsalongwith your schema#1103
Testing done
Did multi turn tool calling with empty string arguments in assistant tool call message
payload for testing
{ "model": "claude-3-5-sonnet-20240620", "max_tokens": 200, "stream": false, "messages": [ { "role": "user", "content": "Based on the temperature and time in California what do you suggest I wear?" }, { "role": "assistant", "content": "Certainly! I'd be happy to provide you with the current temperature and time in San Francisco. To get this information, I'll need to use two separate tools. Let me fetch that data for you.", "tool_calls": [ { "id": "toolu_01BspjPLY4dtHkKGurn8q9A6", "type": "function", "function": { "name": "get_current_temperature", "arguments": "" } }, { "id": "toolu_014jEfKqGbfFvRaKfiauxgPv", "type": "function", "function": { "name": "get_current_time", "arguments": "{\"location\":\"San Francisco, CA\"}" } } ] }, { "role": "tool", "content": "15 degrees", "tool_call_id": "toolu_01BspjPLY4dtHkKGurn8q9A6" }, { "role": "tool", "content": "10 PM", "tool_call_id": "toolu_014jEfKqGbfFvRaKfiauxgPv" } ], "tools": [ { "type": "function", "function": { "name": "get_current_temperature", "description": "Get the current temperature for a specific location", "parameters": { "type": "object", "properties": { "location": { "type": "string", "description": "The city and state, e.g., San Francisco, CA" }, "unit": { "type": "string", "enum": [ "Celsius", "Fahrenheit" ], "description": "The temperature unit to use. Infer this from the user's location." } }, "required": [ "location", "unit" ] } } }, { "type": "function", "function": { "name": "get_current_time", "description": "Get the current time for a specific location", "parameters": { "type": "object", "properties": { "location": { "type": "string", "description": "The city and state, e.g., San Francisco, CA" } }, "required": [ "location" ] } } } ] }