Troubleshooting of Ollama API

Troubleshooting of Ollama API

If you have any questions about the Ollama API, you can post them on the Ollama GitHub Repository.

How to Fix "Unauthorized" Error

Error message:

HTTP Error 401
Response: {"error":"Unauthorized"}

Solution

Verify that your API key is correct by testing it with the following curl command:

curl -s -o /dev/null -w "%{http_code}\n" -X POST https://ollama.com/api/chat -H "Authorization: Bearer REPLACE_WITH_YOUR_API_KEY" -H "Content-Type: application/json" -d '{"model":"gpt-oss:120b","messages":[{"role":"user","content":"hi"}],"stream":false}'

Check the returned status code:

• 200 — the API key is valid
• 401 — the API key is invalid or incorrect

How to fix "prompt too long; exceeded max context length" when sending files

Error message:

HTTP Error 400: {"error":"prompt too long; exceeded max context length by 311181 tokens"}

Applicable Model:

• gpt-oss:120b

Solution

Root Cause: The entire file content was concatenated into content as raw text. Ollama's /api/chat has no dedicated "file" field — text/document content must be extracted client-side and embedded directly into the message string, which counts fully against num_ctx.

Incorrect Usage: Whole file dumped straight into content

'messages' => [[
    'role' => 'user',
    'content' => file''get''contents($path) . "\n\nPlease summary this document"
]]

Correct Usage: Truncate / chunk / summarize before sending, and raise num_ctx to match

'messages' => [[
    'role' => 'user',
    'content' => $chunkedText . "\n\nPlease summary this document"
]],
'options' => [
    'num_ctx' => 32768
]

Explanation: Ollama's chat endpoint only has a native images array (base64-encoded, vision models only). There is no files or document parameter. Any non-image file — PDF, txt, docx, log — must be converted to plain text on the client and inserted into content, so its size is not free; it consumes tokens exactly like typed text.

If the resulting text exceeds the model's context window (num_ctx, default is often small, e.g. 2048–4096), the request fails with "prompt too long". Fix by combining:

• Explicitly setting a larger num_ctx in options (bounded by what the model/hardware can actually support)
• Chunking the file and sending it in multiple requests
• Pre-summarizing or using RAG-style retrieval so only relevant excerpts are sent, instead of the full file

This fix resolves the "prompt too long; exceeded max context length" error when large files are pasted into chat.

How to fix "json: cannot unmarshal object ..."

Error message:

HTTP Error 400: {"error": "json: cannot unmarshal object into Go struct field ChatRequest.messages of type []api.Message"}

Applicable Model:

• gpt-oss:120b

Solution

Root Cause: Incorrect `messages` format. The API expects an array format, but a single object was being sent.

Incorrect Format in PHP: ❌ This is an object

'messages' => ['role' => 'user', 'content' => $message]

Correct Format in PHP: ✅ This is an array containing an object

'messages' => [['role' => 'user', 'content' => $message]]  

Explanation: The `messages` parameter must be an array of message objects, not a single message object. Each message object should contain `role` and `content` properties. The extra square brackets `[]` wrap the message object into an array, which matches the API's expected format.

This fix resolves the `"cannot unmarshal object into Go struct field ChatRequest.messages"` error.

Related links

• Forum: ollama