github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-06 16:11:34 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #9106] Return function result to AI by API. #5928

New Issue
Closed
opened 2026-04-12 17:16:01 -05:00 by GiteaMirror · 6 comments
GiteaMirror commented 2026-04-12 17:16:01 -05:00
Owner
Copy Link

Originally created by @Jauhuei on GitHub (Feb 14, 2025).
Original GitHub issue: https://github.com/ollama/ollama/issues/9106

{
  "model": "llama3.2",
  "created_at": "2024-07-22T20:33:28.123648Z",
  "message": {
    "role": "assistant",
    "content": "",
    "tool_calls": [
      {
        "function": {
          "name": "get_current_weather",
          "arguments": {
            "format": "celsius",
            "location": "Paris, FR"
          }
        }
      }
    ]
  },
  "done_reason": "stop",
  "done": true,
  "total_duration": 885095291,
  "load_duration": 3753500,
  "prompt_eval_count": 122,
  "prompt_eval_duration": 328493000,
  "eval_count": 33,
  "eval_duration": 552222000
}

I executed the get_current_weather function and received the following result:

{
  "city": "Paris, FR",
  "temperature": 20
}

How do I return this result to the AI?

Originally created by @Jauhuei on GitHub (Feb 14, 2025). Original GitHub issue: https://github.com/ollama/ollama/issues/9106 ``` { "model": "llama3.2", "created_at": "2024-07-22T20:33:28.123648Z", "message": { "role": "assistant", "content": "", "tool_calls": [ { "function": { "name": "get_current_weather", "arguments": { "format": "celsius", "location": "Paris, FR" } } } ] }, "done_reason": "stop", "done": true, "total_duration": 885095291, "load_duration": 3753500, "prompt_eval_count": 122, "prompt_eval_duration": 328493000, "eval_count": 33, "eval_duration": 552222000 } ``` I executed the `get_current_weather` function and received the following result: ``` { "city": "Paris, FR", "temperature": 20 } ``` How do I return this result to the AI?
GiteaMirror commented 2026-04-12 17:16:02 -05:00
Author
Owner
Copy Link

@Jauhuei commented on GitHub (Feb 14, 2025):

For example:

I want to use Ollama to create an AI.

When I ask the AI through C# if I can wear a short-sleeve shirt tomorrow, the AI responds by making a Tool_Calls request to check the weather temperature.

C# responds to the AI with the weather temperature.

Then, the AI answers that it's okay to wear a short-sleeve shirt.

<!-- gh-comment-id:2659396608 --> @Jauhuei commented on GitHub (Feb 14, 2025): For example: I want to use Ollama to create an AI. When I ask the AI through C# if I can wear a short-sleeve shirt tomorrow, the AI responds by making a Tool_Calls request to check the weather temperature. C# responds to the AI with the weather temperature. Then, the AI answers that it's okay to wear a short-sleeve shirt.
GiteaMirror commented 2026-04-12 17:16:03 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Feb 14, 2025):

By sending the result back as a tool role:

tool_response = execute_tool();  // {"city": "Paris, FR","temperature": 20}
messages.Add(new Dictionary<string, string>(){
    { "role", "tool" },
    { "content", tool_response }
});

An example here is in python but the logic is the same, the data structure that is used in C# to track the messages just gets the tool call result appended to it and is sent to the LLM to generate the answer to the question.

<!-- gh-comment-id:2659563204 --> @rick-github commented on GitHub (Feb 14, 2025): By sending the result back as a tool role: ```C# tool_response = execute_tool(); // {"city": "Paris, FR","temperature": 20} messages.Add(new Dictionary<string, string>(){ { "role", "tool" }, { "content", tool_response } }); ``` An example [here](https://github.com/ollama/ollama-python/blob/main/examples/tools.py) is in python but the logic is the same, the data structure that is used in C# to track the messages just gets the tool call result appended to it and is sent to the LLM to generate the answer to the question.
GiteaMirror commented 2026-04-12 17:16:03 -05:00
Author
Owner
Copy Link

@Jauhuei commented on GitHub (Feb 15, 2025):

{
    "model": "llama3.2:3b",
    "messages": [
        {
            "content": "What is the temperature in Suzhou and Beijing today?",
            "role": "user"
        }
    ],
    "stream": false,
    "tools": [
        {
            "type": "function",
            "description": "Get the weather information for the specified city",
            "name": "getWeather",
            "parameters": {
                "properties": {
                    "city": {
                        "description": "The name of the city for which to query the weather (only one city name is allowed)",
                        "type": "string"
                    }
                },
                "required": [
                    "city"
                ],
                "type": "object"
            }
        }
    ]
}

The AI didn't call the method according to the format. Regardless of how smart the AI is, I always hope it can at least follow the document format.
The content in the tool response should also follow a more standardized format, so that the AI can understand the results after it makes the call.

{
    "model": "llama3.2:3b",
    "created_at": "2025-02-15T14:15:27.003837Z",
    "message": {
        "role": "assistant",
        "content": "",
        "tool_calls": [
            {
                "function": {
                    "name": "wikipedia.page",
                    "arguments": {
                        "id": "Suzhou, China" // error
                    }
                }
            },
            {
                "function": {
                    "name": "wikipedia.page",
                    "arguments": {
                        "id": "Beijing, China"
                    }
                }
            }
        ]
    },
    "done_reason": "stop",
    "done": true,
    "total_duration": 6565786000,
    "load_duration": 3210631000,
    "prompt_eval_count": 143,
    "prompt_eval_duration": 1822000000,
    "eval_count": 35,
    "eval_duration": 1531000000
}
<!-- gh-comment-id:2660944860 --> @Jauhuei commented on GitHub (Feb 15, 2025): ``` { "model": "llama3.2:3b", "messages": [ { "content": "What is the temperature in Suzhou and Beijing today?", "role": "user" } ], "stream": false, "tools": [ { "type": "function", "description": "Get the weather information for the specified city", "name": "getWeather", "parameters": { "properties": { "city": { "description": "The name of the city for which to query the weather (only one city name is allowed)", "type": "string" } }, "required": [ "city" ], "type": "object" } } ] } ``` The AI didn't call the method according to the format. Regardless of how smart the AI is, I always hope it can at least follow the document format. The content in the tool response should also follow a more standardized format, so that the AI can understand the results after it makes the call. ``` { "model": "llama3.2:3b", "created_at": "2025-02-15T14:15:27.003837Z", "message": { "role": "assistant", "content": "", "tool_calls": [ { "function": { "name": "wikipedia.page", "arguments": { "id": "Suzhou, China" // error } } }, { "function": { "name": "wikipedia.page", "arguments": { "id": "Beijing, China" } } } ] }, "done_reason": "stop", "done": true, "total_duration": 6565786000, "load_duration": 3210631000, "prompt_eval_count": 143, "prompt_eval_duration": 1822000000, "eval_count": 35, "eval_duration": 1531000000 } ```
GiteaMirror commented 2026-04-12 17:16:04 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Feb 15, 2025):

It helps if the tool is formatted a bit better.

$ echo '{
  "model": "llama3.2:3b",
  "messages": [
    {
      "content": "What is the temperature in Suzhou and Beijing today?",
      "role": "user"
    }
  ],
  "stream": false,
  "tools": [
    {
      "type": "function",
      "function": {
        "name": "getWeather",
        "description": "Get the weather for a city",
        "parameters": {
          "type": "object",
          "required": [
            "city"
          ],
          "properties": {
            "city": {
              "type": "string",
              "description": "The name of the city for which to query the weather (only one city name is allowed)"
            }
          }
        }
      }
    }
  ]
}' | curl -s localhost:11434/api/chat -d@- | jq


{
  "model": "llama3.2:3b",
  "created_at": "2025-02-15T16:25:02.199693357Z",
  "message": {
    "role": "assistant",
    "content": "",
    "tool_calls": [
      {
        "function": {
          "name": "getWeather",
          "arguments": {
            "city": "Suzhou"
          }
        }
      },
      {
        "function": {
          "name": "getWeather",
          "arguments": {
            "city": "Beijing"
          }
        }
      }
    ]
  },
  "done_reason": "stop",
  "done": true,
  "total_duration": 460059385,
  "load_duration": 199528465,
  "prompt_eval_count": 175,
  "prompt_eval_duration": 9000000,
  "eval_count": 37,
  "eval_duration": 250000000
}
<!-- gh-comment-id:2660992952 --> @rick-github commented on GitHub (Feb 15, 2025): It helps if the tool is formatted a bit better. ```sh $ echo '{ "model": "llama3.2:3b", "messages": [ { "content": "What is the temperature in Suzhou and Beijing today?", "role": "user" } ], "stream": false, "tools": [ { "type": "function", "function": { "name": "getWeather", "description": "Get the weather for a city", "parameters": { "type": "object", "required": [ "city" ], "properties": { "city": { "type": "string", "description": "The name of the city for which to query the weather (only one city name is allowed)" } } } } } ] }' | curl -s localhost:11434/api/chat -d@- | jq ``` ```console { "model": "llama3.2:3b", "created_at": "2025-02-15T16:25:02.199693357Z", "message": { "role": "assistant", "content": "", "tool_calls": [ { "function": { "name": "getWeather", "arguments": { "city": "Suzhou" } } }, { "function": { "name": "getWeather", "arguments": { "city": "Beijing" } } } ] }, "done_reason": "stop", "done": true, "total_duration": 460059385, "load_duration": 199528465, "prompt_eval_count": 175, "prompt_eval_duration": 9000000, "eval_count": 37, "eval_duration": 250000000 } ```
GiteaMirror commented 2026-04-12 17:16:04 -05:00
Author
Owner
Copy Link

@Jauhuei commented on GitHub (Feb 16, 2025):

@rick-github
Thank you, with your format, AI can indeed successfully query the weather for Suzhou and Beijing. Now, how can I correctly tell it the temperatures for Suzhou and Beijing in two separate responses? According to the previous answerer's suggestion, should I tell the AI like this?

{
    "model": "llama3.2:3b",
    "messages": [
        {
            "content": "What is the temperature in Suzhou and Beijing today?",
            "role": "user"
        },
        {
            "role": "assistant",
            "content": "",
            "tool_calls": [
                {
                    "function": {
                        "name": "getWeather",
                        "arguments": {
                            "city": "Suzhou"
                        }
                    }
                },
                {
                    "function": {
                        "name": "getWeather",
                        "arguments": {
                            "city": "Beijing"
                        }
                    }
                }
            ]
        },
        {
            "role": "tool",
            "content": "{\"city\":\"Suzhou\", \"temperature\":20}",
            "name": "getWeather"
        },
        {
            "role": "tool",
            "content": "{\"city\":\"Beijing\", \"temperature\":15}",
            "name": "getWeather"
        }
    ],
    "stream": false,
    "tools": [
        {
            "type": "function",
            "function": {
                "name": "getWeather",
                "description": "Get the weather for a city",
                "parameters": {
                    "type": "object",
                    "required": [
                        "city"
                    ],
                    "properties": {
                        "city": {
                            "type": "string",
                            "description": "The name of the city for which to query the weather (only one city name is allowed)"
                        }
                    }
                }
            }
        }
    ]
}
<!-- gh-comment-id:2661223537 --> @Jauhuei commented on GitHub (Feb 16, 2025): @rick-github Thank you, with your format, AI can indeed successfully query the weather for Suzhou and Beijing. Now, how can I correctly tell it the temperatures for Suzhou and Beijing in two separate responses? According to the previous answerer's suggestion, should I tell the AI like this? ``` { "model": "llama3.2:3b", "messages": [ { "content": "What is the temperature in Suzhou and Beijing today?", "role": "user" }, { "role": "assistant", "content": "", "tool_calls": [ { "function": { "name": "getWeather", "arguments": { "city": "Suzhou" } } }, { "function": { "name": "getWeather", "arguments": { "city": "Beijing" } } } ] }, { "role": "tool", "content": "{\"city\":\"Suzhou\", \"temperature\":20}", "name": "getWeather" }, { "role": "tool", "content": "{\"city\":\"Beijing\", \"temperature\":15}", "name": "getWeather" } ], "stream": false, "tools": [ { "type": "function", "function": { "name": "getWeather", "description": "Get the weather for a city", "parameters": { "type": "object", "required": [ "city" ], "properties": { "city": { "type": "string", "description": "The name of the city for which to query the weather (only one city name is allowed)" } } } } } ] } ```
GiteaMirror commented 2026-04-12 17:16:05 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Feb 16, 2025):

Yes.

<!-- gh-comment-id:2661439701 --> @rick-github commented on GitHub (Feb 16, 2025): Yes.
Sign in to join this conversation.
No Branch/Tag Specified
Branches Tags
Labels
Clear labels
amd
api
app
bug
build
cli
cloud
compatibility
context-length
create
docker
documentation
embeddings
feature request
feedback wanted
good first issue
gpt-oss
gpu
harmony
help wanted
image
install
intel
js
launch
linux
macos
memory
mlx
model
needs more info
networking
nvidia
ollama.com
performance
pull-request
Mirrored from GitHub Pull Request
 python
question
registry
rendering
thinking
tools
top
vulkan
windows
wsl
No Label
Milestone
No items
No Milestone
Projects
Clear projects
No project
Assignees
Clear assignees
GiteaMirror ninjasurge
No Assignees
1 Participants
Notifications
Due Date
No due date set.
Dependencies

No dependencies set.

Reference: github-starred/ollama#5928
Delete Branch "%!s()"

Deleting a branch is permanent. Although the deleted branch may continue to exist for a short time before it actually gets removed, it CANNOT be undone in most cases. Continue?