github-starred/ollama
2
0
Fork 0
mirror of https://github.com/ollama/ollama.git synced 2026-05-07 00:22:43 -05:00
Code Issues 14.5k Packages Projects Releases 120 Wiki Activity

[GH-ISSUE #15465] fix: recover truncated tool call JSON when max_tokens cuts output #9885

New Issue
Open
opened 2026-04-12 22:44:40 -05:00 by GiteaMirror · 0 comments
GiteaMirror commented 2026-04-12 22:44:40 -05:00
Owner
Copy Link

Originally created by @Rih0z on GitHub (Apr 9, 2026).
Original GitHub issue: https://github.com/ollama/ollama/issues/15465

Problem

When a model generates a large tool call (e.g., write_file with HTML content), the JSON arguments can be truncated if the output hits the token limit. The tool call is silently dropped instead of returning partial arguments.

This causes missing tool calls when:

  • write_file produces large HTML/code content
  • Qwen3/Gemma4 models hit their output token limit mid-JSON

Proposed Fix

Add recovery for truncated tool call JSON so that at least the successfully parsed key-value pairs are returned rather than dropping the entire call.

Reproduction

curl http://localhost:11434/api/chat -d '{
  "model": "qwen3:8b",
  "messages": [{"role": "user", "content": "Write a complete HTML dashboard page"}],
  "tools": [{"type": "function", "function": {"name": "write_file", "parameters": {"type": "object", "properties": {"path": {"type": "string"}, "content": {"type": "string"}}}}}],
  "options": {"num_predict": 2048}
}'

Tool call JSON gets truncated, no tool_calls in response.

Note: PRs #14835/#14915 fix this for model-specific parsers (Qwen3, etc.). This fix targets the generic tools/tools.go parser used when no model-specific parser is available.

Originally created by @Rih0z on GitHub (Apr 9, 2026). Original GitHub issue: https://github.com/ollama/ollama/issues/15465 ## Problem When a model generates a large tool call (e.g., `write_file` with HTML content), the JSON arguments can be truncated if the output hits the token limit. The tool call is silently dropped instead of returning partial arguments. This causes missing tool calls when: - `write_file` produces large HTML/code content - Qwen3/Gemma4 models hit their output token limit mid-JSON ## Proposed Fix Add recovery for truncated tool call JSON so that at least the successfully parsed key-value pairs are returned rather than dropping the entire call. ## Reproduction ```bash curl http://localhost:11434/api/chat -d '{ "model": "qwen3:8b", "messages": [{"role": "user", "content": "Write a complete HTML dashboard page"}], "tools": [{"type": "function", "function": {"name": "write_file", "parameters": {"type": "object", "properties": {"path": {"type": "string"}, "content": {"type": "string"}}}}}], "options": {"num_predict": 2048} }' ``` Tool call JSON gets truncated, no tool_calls in response. Note: PRs #14835/#14915 fix this for model-specific parsers (Qwen3, etc.). This fix targets the generic `tools/tools.go` parser used when no model-specific parser is available.
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#9885
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?