fix(ai-ollama): report finishReason length for truncated Ollama streams

.changeset/ollama-stream-length.md

@@ -0,0 +1,5 @@
+---
+'@tanstack/ai-ollama': patch
+---
+
+Preserve Ollama's `done_reason: "length"` in streamed chat completions, including turns with tool calls, so truncated output does not execute tools.

docs/adapters/ollama.md

@@ -212,6 +212,8 @@ const stream = chat({
 
 Sampling options do not go at the root of `chat()`. For Ollama, they go to `modelOptions.options.temperature`, `modelOptions.options.top_p`, and `modelOptions.options.num_predict`. See [Moving Sampling Options into modelOptions](../migration/sampling-options-to-model-options). The full option list is in [All runner options](#all-runner-options).
 
+When Ollama reports `done_reason: "length"`, the final `RUN_FINISHED` event carries `metadata.tanstack.finishReason: "length"`. This also applies when the stream contains tool calls. `chat()` keeps the partial reply and skips tool execution for that truncated turn.
+
 ### Reasoning
 
 Running a thinking model like Qwen 3 or gpt-oss? Set `reasoning` on `chat()`:

docs/config.json

@@ -1715,7 +1715,7 @@
           "label": "Ollama",
           "to": "adapters/ollama",
           "addedAt": "2026-04-15",
-          "updatedAt": "2026-10-22"
+          "updatedAt": "2026-10-25"
         },
         {
           "label": "Grok (xAI)",

packages/ai-ollama/src/adapters/text.ts

@@ -432,7 +432,12 @@ export class OllamaTextAdapter<TModel extends string> extends BaseTextAdapter<
           threadId,
           model: chunk.model,
           timestamp: Date.now(),
-          finishReason: toolCallsEmitted.size > 0 ? 'tool_calls' : 'stop',
+          finishReason:
+            chunk.done_reason === 'length'
+              ? 'length'
+              : toolCallsEmitted.size > 0
+                ? 'tool_calls'
+                : 'stop',
           // usage is optional under exactOptionalPropertyTypes; omit the key
           // entirely when Ollama reported no token counts.
           ...(finishUsage && { usage: finishUsage }),

packages/ai-ollama/tests/stream-finish-reason.test.ts

@@ -0,0 +1,85 @@
+import { afterEach, describe, expect, it, vi } from 'vitest'
+import { chat } from '@tanstack/ai'
+import { resolveDebugOption } from '@tanstack/ai/adapter-internals'
+import { createOllamaChat } from '../src/adapters/text'
+
+afterEach(() => vi.unstubAllGlobals())
+
+describe('Ollama stream finish reason', () => {
+  it.each([
+    { doneReason: 'length', toolCall: false, expected: 'length' },
+    { doneReason: 'length', toolCall: true, expected: 'length' },
+    { doneReason: 'stop', toolCall: false, expected: 'stop' },
+    { doneReason: 'stop', toolCall: true, expected: 'tool_calls' },
+    { doneReason: undefined, toolCall: false, expected: 'stop' },
+    { doneReason: undefined, toolCall: true, expected: 'tool_calls' },
+  ])(
+    'maps $doneReason with toolCall=$toolCall to $expected',
+    async ({ doneReason, toolCall, expected }) => {
+      const lines = [
+        {
+          message: {
+            role: 'assistant',
+            content: 'Partial answer',
+            ...(toolCall && {
+              tool_calls: [{ function: { name: 'lookup', arguments: {} } }],
+            }),
+          },
+          done: false,
+        },
+        {
+          message: { role: 'assistant', content: '' },
+          done: true,
+          done_reason: doneReason,
+          prompt_eval_count: 14,
+          eval_count: 5,
+        },
+      ]
+      const fetch = vi.fn(
+        async () =>
+          new Response(
+            lines.map((line) => JSON.stringify(line) + '\n').join(''),
+            { headers: { 'content-type': 'application/x-ndjson' } },
+          ),
+      )
+      vi.stubGlobal('fetch', fetch)
+      const adapter = createOllamaChat('llama3.2', 'http://ollama.test')
+      const messages = [{ role: 'user' as const, content: 'Reply' }]
+      const chunks = []
+      for await (const chunk of adapter.chatStream({
+        threadId: 'ollama-finish-reason',
+        logger: resolveDebugOption(false),
+        model: adapter.model,
+        messages,
+      }))
+        chunks.push(chunk)
+
+      expect(chunks.at(-1)).toMatchObject({
+        type: 'RUN_FINISHED',
+        finishReason: expected,
+      })
+
+      if (doneReason === 'length') {
+        const execute = vi.fn(() => 'result')
+        const events = []
+        for await (const event of chat({
+          threadId: 'ollama-finish-reason',
+          adapter,
+          messages,
+          tools: [{ name: 'lookup', description: 'Look up an item', execute }],
+          modelOptions: { model: adapter.model, options: { num_predict: 5 } },
+          debug: false,
+        }))
+          events.push(event)
+
+        expect(events.at(-1)).toMatchObject({
+          type: 'RUN_FINISHED',
+          metadata: { tanstack: { finishReason: 'length' } },
+        })
+        expect(events.some((event) => event.type === 'RUN_ERROR')).toBe(false)
+        expect(execute).not.toHaveBeenCalled()
+        expect(fetch).toHaveBeenCalledTimes(2)
+      }
+    },
+  )
+})