BUG: Anthropic API - `tool_use` ids were found without `tool_result` blocks

[malua]

Hey,

it seems there is an issue with calling the Anthropic API:

2026-02-22T09:32:04.241015Z ERROR handle_message: spacebot::agent::channel: message=channel LLM call failedchannel_id=portal:chat:spacer-1error=CompletionError: ProviderError: Anthropic API error (400 Bad Request): messages.2: tool_use ids were found without tool_result blocks immediately after: toolu_01HDS94VNCHudZHmhBTBRGT7. Each tool_use block must have a corresponding tool_result block in the next message. channel_id=portal:chat:spacer-1agent_id=spacer-1message_id=0b529f03-1df2-428a-8253-d8f3b9071a5e

Thanks for checking it out!

[fnumatic]

Bug: Chat stops responding after first message - regression in v0.1.13

Summary

The error invalid params, tool call result does not follow tool call (2013) is a regression introduced in v0.1.13. The bug does NOT occur in v0.1.12.

Error Log

spacebot  | 2026-02-22T09:03:24.402999Z  INFO handle_message:run_agent_turn: rig::agent::prompt_request: message=Current conversation depth: 1/5
spacebot  | 2026-02-22T09:03:25.434498Z ERROR handle_message: spacebot::agent::channel: message=channel LLM call failederror=CompletionError: ProviderError: OpenAI API error (400 Bad Request): invalid params, tool call result does not follow tool call (2013)

Reproduction

1. Run spacebot v0.1.13 with MiniMax model (minimax/MiniMax-M2.5)
2. Send a message to the agent
3. First message works (or may fail depending on timing)
4. Second message always fails with the tool call error
5. All subsequent messages fail with the same error
6. Only container restart temporarily fixes it

Verified: Works in v0.1.12

Tested and confirmed: v0.1.12 does NOT have this bug. The agent works correctly with the same MiniMax model configuration.

docker-compose.yml:
image: ghcr.io/spacedriveapp/spacebot:v0.1.12-full  # WORKS
image: ghcr.io/spacedriveapp/spacebot:full           # BROKEN (v0.1.13)

Environment

• Broken version: 0.1.13 (ghcr.io/spacedriveapp/spacebot:full)
• Working version: 0.1.12 (ghcr.io/spacedriveapp/spacebot:v0.1.12-full)
• Model: minimax/minimax-M2.5
• API Type: openai_completions

Root Cause

The bug is a regression introduced in v0.1.13. It manifests deterministically - every second message fails with the tool call protocol error.

The history rollback fixes (PR #104, #114) are NOT yet released - they will be in v0.1.14. However, the issue appears to be different from just missing rollback logic, since v0.1.12 works fine.

Changes in v0.1.13 Likely Causing the Regression

1. Prompt Changes (prompts/en/channel.md.j2)

• Modified agent behavior instructions for branching, blocking
• Added new memory types (identity, event, observation)
• Changed response guidelines

2. Channel Handling (src/agent/channel.rs)

• Added reply_target_message_id tracking
• Added branch_reply_targets HashMap
• Changed retry logic syntax (if-let chaining)
• Modified retrigger handling

3. Retry Logic Change

The retry logic was reformatted from nested if-blocks to if-let chaining:

// v0.1.12 (works)
if let Ok(ref response) = result {
    if extract_reply_from_tool_syntax(response.trim()).is_some() {
        // retry logic
    }
}

// v0.1.13 (broken)  
if let Ok(ref response) = result
    && extract_reply_from_tool_syntax(response.trim()).is_some()
{
    // retry logic
}

Proposed Fix

1. Identify the exact change causing the regression
2. Either revert the problematic change OR
3. Backport the history rollback logic from PR Fix: roll back conversation history after LLM hard errors #104/fix(channel): roll back history on PromptCancelled to prevent poisoned turns #114 into v0.1.13

Workaround

Use v0.1.12 until v0.1.14 is released:

# docker-compose.yml
image: ghcr.io/spacedriveapp/spacebot:v0.1.12-full

Related

• PR Fix: roll back conversation history after LLM hard errors #104: "Roll back conversation history after LLM hard errors" (NOT YET RELEASED)
• PR fix(channel): roll back history on PromptCancelled to prevent poisoned turns #114: "fix(channel): roll back history on PromptCancelled" (NOT YET RELEASED)

[dallasrose]

having a similar error: ERROR handle_message: spacebot::agent::channel: message=channel LLM call failedchannel_id=portal:chat:atlaserror=CompletionError: ProviderError: OpenAI API error (400 Bad Request): Invalid request: an assistant message with 'tool_calls' must be followed by tool messages responding to each 'tool_call_id'. The following tool_call_ids did not have response messages: reply:1 channel_id=portal:chat:atlasagent_id=atlasmessage_id=496479e4-db63-40e0-860e-8c55ca9e65a5

[Marenz]

@fnumatic your post is a bit unwieldily and repetitive, please keep it readable for humans, too.
For example , I am not sure if you have tested now with #114 or not.

Did anyone?

[Marenz]

Is this docker image working? Is the bug fixed in there?

ghcr.io/marenz/spacebot:136-test

[malua]

thanks for the test image, but i don't think its publicly available?

[fnumatic]

Not tested with #114 – I'm on v0.1.13 as stated in the issue. The bug appeared after upgrading from v0.1.12 to v0.1.13.

// v0.1.12 (works)
if let Ok(ref response) = result {
    if extract_reply_from_tool_syntax(response.trim()).is_some() {
        // retry logic
    }
}

// v0.1.13 (broken)  
if let Ok(ref response) = result
    && extract_reply_from_tool_syntax(response.trim()).is_some()
{
    // retry logic
}

The error is most likely in the code block shown. This is subtle behavior related to evaluation order in Rust's if-let combined with && operator

regarding ghcr.io/marenz/spacebot:136-test
docker compose pull
[+] Pulling 1/1
✘ spacebot Error Head "https://ghcr.io/v2/marenz/spacebot/manifests/136-test": unauthorized
Error response from daemon: Head "https://ghcr.io/v2/marenz/spacebot/manifests/136-test": unauthorized

[Marenz]

thanks for the test image, but i don't think its publicly available?

Oops, my bad. Rectified.

[malua]

Yes, that works for me! The original error is gone.
But now i get this error every other message or i do not get any response.

2026-02-22T17:46:41.009580Z ERROR handle_message: spacebot::agent::channel: message=channel LLM call failedchannel_id=portal:chat:spacer-1error=CompletionError: ProviderError: anthropic/claude-sonnet-4-6 failed after 3 attempts: ResponseError: empty response from Anthropic channel_id=portal:chat:spacer-1agent_id=spacer-1message_id=0cda00a1-ef35-4717-bab2-d781953dbe02

I am not sure if thats related though... Maybe that's just Anthropics API having some issues. thanks anyway!

2026-02-22T17:45:44.224645Z  INFO spacebot::tools::reply: message=reply tool calledconversation_id=portal:chat:spacer-1content_len=116
2026-02-22T17:45:44.226188Z  INFO spacebot: message=routing outbound response to messaging adapterconversation_id=portal:chat:spacer-1
2026-02-22T17:46:30.106906Z  INFO handle_message: spacebot::agent::channel: message=handling messagechannel_id=portal:chat:spacer-1message_id=0cda00a1-ef35-4717-bab2-d781953dbe02 channel_id=portal:chat:spacer-1agent_id=spacer-1message_id=0cda00a1-ef35-4717-bab2-d781953dbe02
2026-02-22T17:46:30.108695Z  INFO handle_message:run_agent_turn: rig::agent::prompt_request: message=Current conversation depth: 1/5 channel_id=portal:chat:spacer-1agent_id=spacer-1message_id=0cda00a1-ef35-4717-bab2-d781953dbe02 conversation_id="portal:chat:spacer-1"channel_id=portal:chat:spacer-1agent_id=spacer-1
2026-02-22T17:46:32.986679Z  INFO spacebot::tools::skip: message=skip tool called, suppressing responsereason="Message appears to be cut off mid-sentence. User likely hasn't finished typing yet."
2026-02-22T17:46:32.986811Z  INFO handle_message:run_agent_turn:execute_tool: rig::agent::prompt_request: message=executed tool skip with args {"reason":"Message appears to be cut off mid-sentence. User likely hasn't finished typing yet."}. result: {"skipped":true} channel_id=portal:chat:spacer-1agent_id=spacer-1message_id=0cda00a1-ef35-4717-bab2-d781953dbe02 conversation_id="portal:chat:spacer-1"channel_id=portal:chat:spacer-1agent_id=spacer-1 gen_ai.operation.name="execute_tool"gen_ai.tool.type="function" gen_ai.tool.name="skip" gen_ai.tool.call.id="toolu_011ELv9TzSuW7uiigKFvMzwi"  
2026-02-22T17:46:32.986917Z  INFO handle_message:run_agent_turn: rig::agent::prompt_request: message=Current conversation depth: 2/5 channel_id=portal:chat:spacer-1agent_id=spacer-1message_id=0cda00a1-ef35-4717-bab2-d781953dbe02 conversation_id="portal:chat:spacer-1"channel_id=portal:chat:spacer-1agent_id=spacer-1
2026-02-22T17:46:34.811729Z  WARN handle_message:run_agent_turn:chat: spacebot::llm::model: message=retriable errormodel=anthropic/claude-sonnet-4-6attempt=1error=ResponseError: empty response from Anthropic channel_id=portal:chat:spacer-1agent_id=spacer-1message_id=0cda00a1-ef35-4717-bab2-d781953dbe02 conversation_id="portal:chat:spacer-1"channel_id=portal:chat:spacer-1agent_id=spacer-1 gen_ai.operation.name="chat"gen_ai.agent.name="Unnamed Agent"
2026-02-22T17:46:37.163221Z  WARN handle_message:run_agent_turn:chat: spacebot::llm::model: message=retriable errormodel=anthropic/claude-sonnet-4-6attempt=2error=ResponseError: empty response from Anthropic channel_id=portal:chat:spacer-1agent_id=spacer-1message_id=0cda00a1-ef35-4717-bab2-d781953dbe02 conversation_id="portal:chat:spacer-1"channel_id=portal:chat:spacer-1agent_id=spacer-1 gen_ai.operation.name="chat"gen_ai.agent.name="Unnamed Agent"
2026-02-22T17:46:41.009094Z  WARN handle_message:run_agent_turn:chat: spacebot::llm::model: message=retriable errormodel=anthropic/claude-sonnet-4-6attempt=3error=ResponseError: empty response from Anthropic channel_id=portal:chat:spacer-1agent_id=spacer-1message_id=0cda00a1-ef35-4717-bab2-d781953dbe02 conversation_id="portal:chat:spacer-1"channel_id=portal:chat:spacer-1agent_id=spacer-1 gen_ai.operation.name="chat"gen_ai.agent.name="Unnamed Agent"
2026-02-22T17:46:41.009580Z ERROR handle_message: spacebot::agent::channel: message=channel LLM call failedchannel_id=portal:chat:spacer-1error=CompletionError: ProviderError: anthropic/claude-sonnet-4-6 failed after 3 attempts: ResponseError: empty response from Anthropic channel_id=portal:chat:spacer-1agent_id=spacer-1message_id=0cda00a1-ef35-4717-bab2-d781953dbe02
2026-02-22T17:47:23.965959Z  INFO handle_message: spacebot::agent::channel: message=handling messagechannel_id=portal:chat:spacer-1message_id=2d777c55-4000-4ca9-97ca-1a46394bce54 channel_id=portal:chat:spacer-1agent_id=spacer-1message_id=2d777c55-4000-4ca9-97ca-1a46394bce54
2026-02-22T17:47:23.967366Z  INFO handle_message:run_agent_turn: rig::agent::prompt_request: message=Current conversation depth: 1/5 channel_id=portal:chat:spacer-1agent_id=spacer-1message_id=2d777c55-4000-4ca9-97ca-1a46394bce54 conversation_id="portal:chat:spacer-1"channel_id=portal:chat:spacer-1agent_id=spacer-1
2026-02-22T17:47:27.317114Z  INFO spacebot::tools::reply: message=reply tool calledconversation_id=portal:chat:spacer-1content_len=52
2026-02-22T17:47:27.318877Z  INFO spacebot: message=routing outbound response to messaging adapterconversation_id=portal:chat:spacer-1
2026-02-22T17:47:38.839645Z  INFO handle_message: spacebot::agent::channel: message=handling messagechannel_id=portal:chat:spacer-1message_id=22864a03-f71f-4043-af28-9e86283cd5aa channel_id=portal:chat:spacer-1agent_id=spacer-1message_id=22864a03-f71f-4043-af28-9e86283cd5aa
2026-02-22T17:47:38.841162Z  INFO handle_message:run_agent_turn: rig::agent::prompt_request: message=Current conversation depth: 1/5 channel_id=portal:chat:spacer-1agent_id=spacer-1message_id=22864a03-f71f-4043-af28-9e86283cd5aa conversation_id="portal:chat:spacer-1"channel_id=portal:chat:spacer-1agent_id=spacer-1
2026-02-22T17:47:41.442869Z  INFO spacebot::tools::skip: message=skip tool called, suppressing responsereason="Message seems incomplete or directed at someone else - \"not really i just got:\" trails off without a question or request for me."
2026-02-22T17:47:41.442992Z  INFO handle_message:run_agent_turn:execute_tool: rig::agent::prompt_request: message=executed tool skip with args {"reason":"Message seems incomplete or directed at someone else - \"not really i just got:\" trails off without a question or request for me."}. result: {"skipped":true} channel_id=portal:chat:spacer-1agent_id=spacer-1message_id=22864a03-f71f-4043-af28-9e86283cd5aa conversation_id="portal:chat:spacer-1"channel_id=portal:chat:spacer-1agent_id=spacer-1 gen_ai.operation.name="execute_tool"gen_ai.tool.type="function" gen_ai.tool.name="skip" gen_ai.tool.call.id="toolu_012wsXosGWdBmuzzqob8wR7W"  
2026-02-22T17:47:41.443220Z  INFO handle_message:run_agent_turn: rig::agent::prompt_request: message=Current conversation depth: 2/5 channel_id=portal:chat:spacer-1agent_id=spacer-1message_id=22864a03-f71f-4043-af28-9e86283cd5aa conversation_id="portal:chat:spacer-1"channel_id=portal:chat:spacer-1agent_id=spacer-1
2026-02-22T17:47:42.756748Z  WARN handle_message:run_agent_turn:chat: spacebot::llm::model: message=retriable errormodel=anthropic/claude-sonnet-4-6attempt=1error=ResponseError: empty response from Anthropic channel_id=portal:chat:spacer-1agent_id=spacer-1message_id=22864a03-f71f-4043-af28-9e86283cd5aa conversation_id="portal:chat:spacer-1"channel_id=portal:chat:spacer-1agent_id=spacer-1 gen_ai.operation.name="chat"gen_ai.agent.name="Unnamed Agent"
2026-02-22T17:47:44.314346Z  WARN handle_message:run_agent_turn:chat: spacebot::llm::model: message=retriable errormodel=anthropic/claude-sonnet-4-6attempt=2error=ResponseError: empty response from Anthropic channel_id=portal:chat:spacer-1agent_id=spacer-1message_id=22864a03-f71f-4043-af28-9e86283cd5aa conversation_id="portal:chat:spacer-1"channel_id=portal:chat:spacer-1agent_id=spacer-1 gen_ai.operation.name="chat"gen_ai.agent.name="Unnamed Agent"
2026-02-22T17:47:46.365027Z  WARN handle_message:run_agent_turn:chat: spacebot::llm::model: message=retriable errormodel=anthropic/claude-sonnet-4-6attempt=3error=ResponseError: empty response from Anthropic channel_id=portal:chat:spacer-1agent_id=spacer-1message_id=22864a03-f71f-4043-af28-9e86283cd5aa conversation_id="portal:chat:spacer-1"channel_id=portal:chat:spacer-1agent_id=spacer-1 gen_ai.operation.name="chat"gen_ai.agent.name="Unnamed Agent"
2026-02-22T17:47:46.365528Z ERROR handle_message: spacebot::agent::channel: message=channel LLM call failedchannel_id=portal:chat:spacer-1error=CompletionError: ProviderError: anthropic/claude-sonnet-4-6 failed after 3 attempts: ResponseError: empty response from Anthropic channel_id=portal:chat:spacer-1agent_id=spacer-1message_id=22864a03-f71f-4043-af28-9e86283cd5aa
2026-02-22T17:47:54.293790Z  INFO handle_message: spacebot::agent::channel: message=handling messagechannel_id=portal:chat:spacer-1message_id=cc711295-ed61-405a-bb22-1927d15fbe26 channel_id=portal:chat:spacer-1agent_id=spacer-1message_id=cc711295-ed61-405a-bb22-1927d15fbe26
2026-02-22T17:47:54.295485Z  INFO handle_message:run_agent_turn: rig::agent::prompt_request: message=Current conversation depth: 1/5 channel_id=portal:chat:spacer-1agent_id=spacer-1message_id=cc711295-ed61-405a-bb22-1927d15fbe26 conversation_id="portal:chat:spacer-1"channel_id=portal:chat:spacer-1agent_id=spacer-1
2026-02-22T17:48:09.751190Z  INFO spacebot::tools::reply: message=reply tool calledconversation_id=portal:chat:spacer-1content_len=151
2026-02-22T17:48:09.752892Z  INFO spacebot: message=routing outbound response to messaging adapterconversation_id=portal:chat:spacer-1
2026-02-22T17:49:18.047753Z  INFO handle_message: spacebot::agent::channel: message=handling messagechannel_id=portal:chat:spacer-1message_id=0c6faca3-e322-46c1-aa06-3c3248166c86 channel_id=portal:chat:spacer-1agent_id=spacer-1message_id=0c6faca3-e322-46c1-aa06-3c3248166c86
2026-02-22T17:49:18.049326Z  INFO handle_message:run_agent_turn: rig::agent::prompt_request: message=Current conversation depth: 1/5 channel_id=portal:chat:spacer-1agent_id=spacer-1message_id=0c6faca3-e322-46c1-aa06-3c3248166c86 conversation_id="portal:chat:spacer-1"channel_id=portal:chat:spacer-1agent_id=spacer-1
2026-02-22T17:49:20.789129Z  INFO spacebot::tools::reply: message=reply tool calledconversation_id=portal:chat:spacer-1content_len=43
2026-02-22T17:49:20.790915Z  INFO spacebot: message=routing outbound response to messaging adapterconversation_id=portal:chat:spacer-1
2026-02-22T17:49:34.296621Z  INFO handle_message: spacebot::agent::channel: message=handling messagechannel_id=portal:chat:spacer-1message_id=782a8a13-fe59-4bd6-a02b-ce44ad6b17c4 channel_id=portal:chat:spacer-1agent_id=spacer-1message_id=782a8a13-fe59-4bd6-a02b-ce44ad6b17c4
2026-02-22T17:49:34.298294Z  INFO handle_message:run_agent_turn: rig::agent::prompt_request: message=Current conversation depth: 1/5 channel_id=portal:chat:spacer-1agent_id=spacer-1message_id=782a8a13-fe59-4bd6-a02b-ce44ad6b17c4 conversation_id="portal:chat:spacer-1"channel_id=portal:chat:spacer-1agent_id=spacer-1
2026-02-22T17:49:54.638675Z  INFO spacebot::tools::reply: message=reply tool calledconversation_id=portal:chat:spacer-1content_len=184
2026-02-22T17:49:54.640349Z  INFO spacebot: message=routing outbound response to messaging adapterconversation_id=portal:chat:spacer-1

[Marenz]

Yeah, I am not sure about those eirther, I have them too. When you expand the logging its mostly "thinking" responses without content 🤷 we could print them, but .. do we want to? We maybe should log them.

More importantly though: This shows the bug is fixed in main.
Closed in #104

[fnumatic]

Fixed. Error is gone. Thanks

I don't see any other errors in the logs. I'm using Minimax.
Is this just your private branch or merged commit 3444aeb on main?