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 #7572] OpenAI API tool calling doesn't work #66881

New Issue
Closed
opened 2026-05-04 08:35:52 -05:00 by GiteaMirror · 8 comments
GiteaMirror commented 2026-05-04 08:35:52 -05:00
Owner
Copy Link

Originally created by @SinanAkkoyun on GitHub (Nov 8, 2024).
Original GitHub issue: https://github.com/ollama/ollama/issues/7572

What is the issue?

Then OpenAI API endpoint does not support proper tool calling, it just outputs the tool in plaintext as message content: {"name": "get_time", "parameters": {"timezone": "Europe/Berlin"}}

OS

Linux

GPU

Nvidia

CPU

Intel

Ollama version

0.4.0

Originally created by @SinanAkkoyun on GitHub (Nov 8, 2024). Original GitHub issue: https://github.com/ollama/ollama/issues/7572 ### What is the issue? Then OpenAI API endpoint does not support proper tool calling, it just outputs the tool in plaintext as message content: `{"name": "get_time", "parameters": {"timezone": "Europe/Berlin"}}` ### OS Linux ### GPU Nvidia ### CPU Intel ### Ollama version 0.4.0
GiteaMirror added the bugapi labels 2026-05-04 08:35:52 -05:00
GiteaMirror commented 2026-05-04 08:35:55 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Nov 8, 2024):

curl -s http://localhost:11434/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{"model":"llama3.1:8b-instruct-q4_K_M",
  "messages": [
    {
      "role": "user",
      "content": "What is the weather like in Boston and New York?"
    }
  ],
  "tools": [
    {
      "type": "function",
      "function": {
        "name": "get_current_weather",
        "description": "Get the current weather in a given location",
        "parameters": {
          "type": "object",
          "properties": {
            "location": {
              "type": "string",
              "description": "The city and state, e.g. San Francisco, CA"
            },
            "unit": {
              "type": "string",
              "enum": ["celsius", "fahrenheit"]
            }
          },
          "required": ["location"]
        }
      }
    }
  ],
  "tool_choice": "auto",
  "stream":false
}' | jq

{
  "id": "chatcmpl-211",
  "object": "chat.completion",
  "created": 1731063537,
  "model": "llama3.1:8b-instruct-q4_K_M",
  "system_fingerprint": "fp_ollama",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "",
        "tool_calls": [
          {
            "id": "call_jd2vds65",
            "type": "function",
            "function": {
              "name": "get_current_weather",
              "arguments": "{\"location\":\"Boston, MA\",\"unit\":\"fahrenheit\"}"
            }
          },
          {
            "id": "call_f11n0cfo",
            "type": "function",
            "function": {
              "name": "get_current_weather",
              "arguments": "{\"location\":\"New York, NY\",\"unit\":\"celsius\"}"
            }
          }
        ]
      },
      "finish_reason": "tool_calls"
    }
  ],
  "usage": {
    "prompt_tokens": 186,
    "completion_tokens": 55,
    "total_tokens": 241
  }
}
<!-- gh-comment-id:2464411883 --> @rick-github commented on GitHub (Nov 8, 2024): ``` curl -s http://localhost:11434/v1/chat/completions \ -H "Content-Type: application/json" \ -d '{"model":"llama3.1:8b-instruct-q4_K_M", "messages": [ { "role": "user", "content": "What is the weather like in Boston and New York?" } ], "tools": [ { "type": "function", "function": { "name": "get_current_weather", "description": "Get the current weather in a given location", "parameters": { "type": "object", "properties": { "location": { "type": "string", "description": "The city and state, e.g. San Francisco, CA" }, "unit": { "type": "string", "enum": ["celsius", "fahrenheit"] } }, "required": ["location"] } } } ], "tool_choice": "auto", "stream":false }' | jq ``` ```json { "id": "chatcmpl-211", "object": "chat.completion", "created": 1731063537, "model": "llama3.1:8b-instruct-q4_K_M", "system_fingerprint": "fp_ollama", "choices": [ { "index": 0, "message": { "role": "assistant", "content": "", "tool_calls": [ { "id": "call_jd2vds65", "type": "function", "function": { "name": "get_current_weather", "arguments": "{\"location\":\"Boston, MA\",\"unit\":\"fahrenheit\"}" } }, { "id": "call_f11n0cfo", "type": "function", "function": { "name": "get_current_weather", "arguments": "{\"location\":\"New York, NY\",\"unit\":\"celsius\"}" } } ] }, "finish_reason": "tool_calls" } ], "usage": { "prompt_tokens": 186, "completion_tokens": 55, "total_tokens": 241 } } ```
GiteaMirror commented 2026-05-04 08:35:56 -05:00
Author
Owner
Copy Link

@SinanAkkoyun commented on GitHub (Nov 8, 2024):

@rick-github
Sorry, I forgot to say:

  • static doesn't work with Llama3.2
  • streaming doesn't work with Llama3.1
<!-- gh-comment-id:2465785220 --> @SinanAkkoyun commented on GitHub (Nov 8, 2024): @rick-github Sorry, I forgot to say: - static doesn't work with Llama3.2 - streaming doesn't work with Llama3.1
GiteaMirror commented 2026-05-04 08:35:56 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Nov 8, 2024):

#5796

By static, do you mean non-streaming?

If you can supply some examples or logs, it will be much easier to resolve your issue.

<!-- gh-comment-id:2465810679 --> @rick-github commented on GitHub (Nov 8, 2024): #5796 By static, do you mean non-streaming? If you can supply some examples or logs, it will be much easier to resolve your issue.
GiteaMirror commented 2026-05-04 08:35:58 -05:00
Author
Owner
Copy Link

@SinanAkkoyun commented on GitHub (Nov 8, 2024):

Sure thing! Just passing your exact prompt (non-streaming and yes I meant that with static) to Llama3.2 results in a response without tool_call, it just brings the output into a normal chat message. Perhaps because of the pythonic call? I can't tell why

(Sorry I also don't know where to get the correct logs from)

<!-- gh-comment-id:2465843175 --> @SinanAkkoyun commented on GitHub (Nov 8, 2024): Sure thing! Just passing your exact prompt (non-streaming and yes I meant that with static) to Llama3.2 results in a response without tool_call, it just brings the output into a normal chat message. Perhaps because of the pythonic call? I can't tell why (Sorry I also don't know where to get the correct logs from)
GiteaMirror commented 2026-05-04 08:35:59 -05:00
Author
Owner
Copy Link

@SinanAkkoyun commented on GitHub (Nov 8, 2024):

Does the ollama openai API perchance support special tokens in the response somehow?

<!-- gh-comment-id:2465845938 --> @SinanAkkoyun commented on GitHub (Nov 8, 2024): Does the ollama openai API perchance support special tokens in the response somehow?
GiteaMirror commented 2026-05-04 08:36:00 -05:00
Author
Owner
Copy Link

@rick-github commented on GitHub (Nov 8, 2024):

Could you supply the script that you are using to call the server? If I feed my example to llama3.2, I get a tool call:

curl -s http://localhost:11434/v1/chat/completions \
-H "Content-Type: application/json" \
-d '{"model":"llama3.2:3b-instruct-q4_K_M",
  "messages": [
    {
      "role": "user",
      "content": "What is the weather like in Boston and New York?"
    }
  ],
  "tools": [
    {
      "type": "function",
      "function": {
        "name": "get_current_weather",
        "description": "Get the current weather in a given location",
        "parameters": {
          "type": "object",
          "properties": {
            "location": {
              "type": "string",
              "description": "The city and state, e.g. San Francisco, CA"
            },
            "unit": {
              "type": "string",
              "enum": ["celsius", "fahrenheit"]
            }
          },
          "required": ["location"]
        }
      }
    }
  ],
  "tool_choice": "auto",
  "stream":false
}' | jq

{
  "id": "chatcmpl-168",
  "object": "chat.completion",
  "created": 1731105942,
  "model": "llama3.2:3b-instruct-q4_K_M",
  "system_fingerprint": "fp_ollama",
  "choices": [
    {
      "index": 0,
      "message": {
        "role": "assistant",
        "content": "",
        "tool_calls": [
          {
            "id": "call_vxmlccki",
            "type": "function",
            "function": {
              "name": "get_current_weather",
              "arguments": "{\"location\":\"Boston, MA\",\"unit\":\"celsius\"}"
            }
          },
          {
            "id": "call_abyfceyx",
            "type": "function",
            "function": {
              "name": "get_current_weather",
              "arguments": "{\"location\":\"New York, NY\",\"unit\":\"celsius\"}"
            }
          }
        ]
      },
      "finish_reason": "tool_calls"
    }
  ],
  "usage": {
    "prompt_tokens": 195,
    "completion_tokens": 55,
    "total_tokens": 250
  }
}

If you did the usual curl|sh installation, logs can be retrieved with journalctl -u ollama --no-pager. See here for other methods.

Special tokens (like <|user|>) are defined on a model-by-model basis, and are processed through template expansion. There are no special tokens in the generic API handling.

<!-- gh-comment-id:2465868360 --> @rick-github commented on GitHub (Nov 8, 2024): Could you supply the script that you are using to call the server? If I feed my example to llama3.2, I get a tool call: ```console curl -s http://localhost:11434/v1/chat/completions \ -H "Content-Type: application/json" \ -d '{"model":"llama3.2:3b-instruct-q4_K_M", "messages": [ { "role": "user", "content": "What is the weather like in Boston and New York?" } ], "tools": [ { "type": "function", "function": { "name": "get_current_weather", "description": "Get the current weather in a given location", "parameters": { "type": "object", "properties": { "location": { "type": "string", "description": "The city and state, e.g. San Francisco, CA" }, "unit": { "type": "string", "enum": ["celsius", "fahrenheit"] } }, "required": ["location"] } } } ], "tool_choice": "auto", "stream":false }' | jq ``` ```json { "id": "chatcmpl-168", "object": "chat.completion", "created": 1731105942, "model": "llama3.2:3b-instruct-q4_K_M", "system_fingerprint": "fp_ollama", "choices": [ { "index": 0, "message": { "role": "assistant", "content": "", "tool_calls": [ { "id": "call_vxmlccki", "type": "function", "function": { "name": "get_current_weather", "arguments": "{\"location\":\"Boston, MA\",\"unit\":\"celsius\"}" } }, { "id": "call_abyfceyx", "type": "function", "function": { "name": "get_current_weather", "arguments": "{\"location\":\"New York, NY\",\"unit\":\"celsius\"}" } } ] }, "finish_reason": "tool_calls" } ], "usage": { "prompt_tokens": 195, "completion_tokens": 55, "total_tokens": 250 } } ``` If you did the usual `curl|sh` installation, logs can be retrieved with `journalctl -u ollama --no-pager`. See [here](https://github.com/ollama/ollama/blob/main/docs/troubleshooting.md#how-to-troubleshoot-issues) for other methods. Special tokens (like `<|user|>`) are defined on a model-by-model basis, and are processed through template expansion. There are no special tokens in the generic API handling.
GiteaMirror commented 2026-05-04 08:36:00 -05:00
Author
Owner
Copy Link

@SinanAkkoyun commented on GitHub (Nov 8, 2024):

That's super strange, I just copied your curl, I will delete and repull 3.2

<!-- gh-comment-id:2465873487 --> @SinanAkkoyun commented on GitHub (Nov 8, 2024): That's super strange, I just copied your curl, I will delete and repull 3.2
GiteaMirror commented 2026-05-04 08:36:01 -05:00
Author
Owner
Copy Link

@SinanAkkoyun commented on GitHub (Nov 8, 2024):

Yup, I had some old llama3.2 model, thanks!

<!-- gh-comment-id:2465876944 --> @SinanAkkoyun commented on GitHub (Nov 8, 2024): Yup, I had some old llama3.2 model, thanks!
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 api bug
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#66881
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?