CodeCompanionChat (with Anthropic) stuck after response, does not complete or add a "## Me" line

[fnune]

Hi! I'm restraining from posting an issue because I have not been able to reproduce this with the minimal.lua example using Anthropic as an adapter.

Using CodeCompanionChat, I get a proper response, but when it ends, CodeCompanion does not apply the changes, and it does not create a new ## Me heading for me to keep typing. I can, however, press Enter to get a new response. Note that this used to work for me just yesterday.

Screencast_20250128_110339.webm

I have tried resetting my config to what it was a few days ago when things worked. I have also downgraded CodeCompanion version by version back to v11.12.0 but the problem persists.

Here is my healthcheck:

codecompanion: require("codecompanion.health").check()

- Neovim version: 0.10.3
- Log file: /home/fausto/.local/state/nvim/codecompanion.log

Dependencies: ~
- OK plenary.nvim installed
- OK nvim-treesitter installed

Tree-sitter parsers: ~
- OK markdown parser installed
- OK yaml parser installed

Libraries: ~
- OK curl installed

The log file is empty.

I would welcome any ideas as to what might be causing this. Thank you!

[olimorris]

That is very odd. I've never seen that occur before. I'd advise turning the logging to DEBUG or TRACE and checking the log file

[fnune]

Alright, I managed to get some logs. I got confused by the nested opts because I use Lazy to manage plugins and I was going opts = { log_level = "TRACE" } but I should have gone opts = { opts = { log_level = "TRACE" } } on my config. My bad.

Here is the log file. I've redacted some bits and removed some triple backslashes.

[TRACE] 2025-01-29 11:18:41
Chat created with ID 7942063
[TRACE] 2025-01-29 11:18:41
Chat opened with ID 7942063
[TRACE] 2025-01-29 11:18:41
Setting the header for the chat buffer
[DEBUG] 2025-01-29 11:18:47
Calling tool: strategies.chat.tools.editor
[DEBUG] 2025-01-29 11:18:47
Variable found: buffer
[TRACE] 2025-01-29 11:18:47
Calling variable: strategies.chat.variables.buffer
[TRACE] 2025-01-29 11:18:47
Buffer Variable:
---
Buffer ID: 1
Name: redacted.py
Path: /home/fausto/REDACTED
Filetype: python
Content:

# REDACTED: correct contents of the file


[DEBUG] 2025-01-29 11:18:47
Settings:
{
  env = {
    api_key = "ANTHROPIC_API_KEY"
  },
  features = {
    text = true,
    tokens = true,
    vision = true
  },
  handlers = {
    chat_output = <function 1>,
    form_messages = <function 2>,
    form_parameters = <function 3>,
    inline_output = <function 4>,
    on_exit = <function 5>,
    tokens = <function 6>
  },
  headers = {
    ["anthropic-beta"] = "prompt-caching-2024-07-31",
    ["anthropic-version"] = "2023-06-01",
    ["content-type"] = "application/json",
    ["x-api-key"] = "${api_key}"
  },
  name = "anthropic",
  opts = {
    cache_breakpoints = 4,
    cache_over = 300,
    stream = true
  },
  parameters = {
    max_tokens = 4096,
    model = "claude-3-5-sonnet-20241022",
    stream = true,
    temperature = 0
  },
  roles = {
    llm = "assistant",
    user = "user"
  },
  schema = {
    max_tokens = {
      default = 4096,
      desc = "The maximum number of tokens to generate before stopping. This parameter only specifies the absolute maximum number of tokens to generate. Different models have different maximum values for this parameter.",
      mapping = "parameters",
      optional = true,
      order = 2,
      type = "number",
      validate = <function 7>
    },
    model = {
      choices = { "claude-3-5-sonnet-20241022", "claude-3-5-haiku-20241022", "claude-3-opus-20240229", "claude-2.1" },
      default = "claude-3-5-sonnet-20241022",
      desc = "The model that will complete your prompt. See https://docs.anthropic.com/claude/docs/models-overview for additional details and options.",
      mapping = "parameters",
      order = 1,
      type = "enum"
    },
    stop_sequences = {
      desc = "Sequences where the API will stop generating further tokens",
      mapping = "parameters",
      optional = true,
      order = 6,
      subtype = {
        type = "string"
      },
      type = "list",
      validate = <function 8>
    },
    temperature = {
      default = 0,
      desc = "Amount of randomness injected into the response. Ranges from 0.0 to 1.0. Use temperature closer to 0.0 for analytical / multiple choice, and closer to 1.0 for creative and generative tasks. Note that even with temperature of 0.0, the results will not be fully deterministic.",
      mapping = "parameters",
      optional = true,
      order = 3,
      type = "number",
      validate = <function 9>
    },
    top_k = {
      desc = "Only sample from the top K options for each subsequent token. Use top_k to remove long tail low probability responses",
      mapping = "parameters",
      optional = true,
      order = 5,
      type = "number",
      validate = <function 10>
    },
    top_p = {
      desc = "Computes the cumulative distribution over all the options for each subsequent token in decreasing probability order and cuts it off once it reaches a particular probability specified by top_p",
      mapping = "parameters",
      optional = true,
      order = 4,
      type = "number",
      validate = <function 11>
    }
  },
  url = "https://api.anthropic.com/v1/messages",
  <metatable> = {
    __index = {
      extend = <function 12>,
      get_default_settings = <function 13>,
      get_env_vars = <function 14>,
      make_safe = <function 15>,
      map_roles = <function 16>,
      map_schema_to_params = <function 17>,
      new = <function 18>,
      resolve = <function 19>,
      set_env_vars = <function 20>
    }
  }
}
[DEBUG] 2025-01-29 11:18:47
Messages:
{ {
    content = "You are an AI programming assistant named \"CodeCompanion\".\nYou are currently plugged in to the Neovim text editor on a user's machine.\n\nYour core tasks include:\n- Answering general programming questions.\n- Explaining how the code in a Neovim buffer works.\n- Reviewing the selected code in a Neovim buffer.\n- Generating unit tests for the selected code.\n- Proposing fixes for problems in the selected code.\n- Scaffolding code for a new workspace.\n- Finding relevant code to the user's query.\n- Proposing fixes for test failures.\n- Answering questions about Neovim.\n- Running tools.\n\nYou must:\n- Follow the user's requirements carefully and to the letter.\n- Keep your answers short and impersonal, especially if the user responds with context outside of your tasks.\n- Minimize other prose.\n- Use Markdown formatting in your answers.\n- Include the programming language name at the start of the Markdown code blocks.\n- Avoid including line numbers in code blocks.\n- Avoid wrapping the whole response in triple backticks.\n- Only return code that's relevant to the task at hand. You may not need to return all of the code that the user has shared.\n- Use actual line breaks instead of '\\n' in your response to begin new lines.\n- Use '\\n' only when you want a literal backslash followed by a character 'n'.\n- All non-code responses must be in English.\n\nWhen given a task:\n1. Think step-by-step and describe your plan for what to build in pseudocode, written out in great detail, unless asked not to do so.\n2. Output the code in a single code block, being careful to only return relevant code.\n3. You should always generate short suggestions for the next user turns that are relevant to the conversation.\n4. You can only give one reply for each conversation turn.",
    cycle = 1,
    id = 1628281114,
    opts = {
      visible = false
    },
    role = "system"
  }, {
    content = "Add a supershort comment to this file",
    cycle = 1,
    id = 805249948,
    opts = {
      visible = true
    },
    role = "user"
  }, {
    content = "## Tools\n\nYou now have access to tools:\n- These enable you to assist the user with specific tasks\n- The user will outline which specific tools you have access to\n- You trigger a tool by following a specific XML schema which is defined for each tool\n\nYou must:\n- Only use the tool when prompted by the user, despite having access to it\n- Follow the specific tool's schema\n- Respond with the schema in XML format\n- Ensure the schema is in a markdown code block that is designated as XML\n- Ensure any output you're intending to execute will be able to parsed as valid XML\n\nPoints to note:\n- The user detects that you've triggered a tool by using Tree-sitter to parse your markdown response\n- If you call multiple tools within the same response:\n  - Each unique tool MUST be called in its own, individual, XML codeblock\n  - Tools of the same type SHOULD be called in the same XML codeblock\n- If your response doesn't follow the tool's schema, the tool will not execute\n- Tools should not alter your core tasks and how you respond to a user",
    cycle = 1,
    id = 1255510772,
    opts = {
      reference = "tool_system_prompt",
      tag = "tool",
      visible = false
    },
    role = "system"
  }, {
    content = "### Editor Tool\n\n1. **Purpose**: Update the contents of a Neovim buffer\n\n2. **Usage**: Return an XML markdown code block for add, update, or delete operations.\n\n3. **Key Points**:\n  - **Only use when prompted** by user (e.g., \"can you update the code?\", \"update the buffer...\")\n  - Ensure XML is **valid and follows the schema**\n  - **Include indentation** in your code\n  - **Don't escape** special characters\n  - **Wrap code in a CDATA block**, the code could contain characters reserved by XML\n  - Make sure the tools xml block is **surrounded by xml**\n\n4. **Actions**:\n\na) Add:\n\nxml\n<tools>\n  <tool name=\"editor\">\n    <action type=\"add\">\n      <line>203</line>\n      <code><![CDATA[    print('Hello World')]]></code>\n      <buffer>1</buffer>\n    </action>\n  </tool>\n</tools>\n\n\nb) Update:\n\nxml\n<tools>\n  <tool name=\"editor\">\n    <action type=\"update\">\n      <end_line>199</end_line>\n      <buffer>10</buffer>\n      <code><![CDATA[   function M.capitalize()]]></code>\n      <start_line>199</start_line>\n    </action>\n  </tool>\n</tools>\n\n\nc) Delete:\n\nxml\n<tools>\n  <tool name=\"editor\">\n    <action type=\"delete\">\n      <end_line>15</end_line>\n      <start_line>10</start_line>\n      <buffer>14</buffer>\n    </action>\n  </tool>\n</tools>\n\n\n5. **Multiple Actions**: Combine actions in one response if needed:\n\nxml\n<tools>\n  <tool name=\"editor\">\n    <action type=\"delete\">\n      <end_line>13</end_line>\n      <start_line>13</start_line>\n      <buffer>5</buffer>\n    </action>\n    <action type=\"add\">\n      <line>20</line>\n      <code><![CDATA[function M.hello_world()]]></code>\n      <buffer>5</buffer>\n    </action>\n  </tool>\n</tools>\n\n\n6. **Note**:\n  - The tool can handle multiple actions in one response such as adding and deleting code\n  - For the delete action, the <start_line> and <end_line> tags are inclusive\n  - The update action first deletes the range in <start_line> and <end_line> (inclusively) and then adds new code from the <start_line>\n  - If you need to update a single line, set <start_line> and <end_line> to the same value in the update action\n  - Remember you can have multiple update actions after one another.\n  - Account for comment blocks and indentation in your code\n  - If the user supplies no context, it can be assumed that they would like you to update the buffer with the code from your last response\n\nRemember: Minimize explanations unless prompted. Focus on generating correct XML.",
    cycle = 1,
    id = 522137796,
    opts = {
      reference = "<tool>editor</tool>",
      tag = "tool",
      visible = false
    },
    role = "system"
  }, {
    content = "REDACTED",
    cycle = 1,
    id = 542986939,
    opts = {
      reference = "REDACTED",
      tag = "variable",
      visible = false
    },
    role = "user"
  } }
[INFO] 2025-01-29 11:18:47
Chat request started
[INFO] 2025-01-29 11:18:47
ELAPSED TIME: 21.770118
[TRACE] 2025-01-29 11:18:47
Fetching environment variable: ANTHROPIC_API_KEY
[INFO] 2025-01-29 11:18:47
Request body file: /tmp/nix-shell.GEhBos/nvim.fausto/cGRs7Q/1.json
[DEBUG] 2025-01-29 11:18:47
Request:
{ "-sSL", "-D", "/run/user/1001/plenary_curl_b9c14a5b.headers", "-X", "POST", "-H", "Anthropic-Beta: prompt-caching-2024-07-31", "-H", "Content-Type: application/json", "-H", "Anthropic-Version: 2023-06-01", "-H", "X-Api-Key: REDACTED", "-d", "@/tmp/nix-shell.GEhBos/nvim.fausto/cGRs7Q/1.json", "--retry", "3", "--retry-delay", "1", "--keepalive-time", "60", "--connect-timeout", "10", "--tcp-nodelay", "--no-buffer", "https://api.anthropic.com/v1/messages" }
[TRACE] 2025-01-29 11:18:51
Output data:
event: message_start
[TRACE] 2025-01-29 11:18:51
Output data:
data: {"type":"message_start","message":{"id":"msg_01WBS2rQfwopGEwoCK67bqpm","type":"message","role":"assistant","model":"claude-3-5-sonnet-20241022","content":[],"stop_reason":null,"stop_sequence":null,"usage":{"input_tokens":3,"cache_creation_input_tokens":13002,"cache_read_input_tokens":0,"output_tokens":1}}      }
[TRACE] 2025-01-29 11:18:51
Output data:
event: content_block_start
[TRACE] 2025-01-29 11:18:51
Output data:
data: {"type":"content_block_start","index":0,"content_block":{"type":"text","text":""}      }
[TRACE] 2025-01-29 11:18:51
Output data:
event: ping
[TRACE] 2025-01-29 11:18:51
Output data:
data: {"type": "ping"}
[TRACE] 2025-01-29 11:18:51
Output data:
event: content_block_delta
[TRACE] 2025-01-29 11:18:51
Output data:
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"I"}       }
[TRACE] 2025-01-29 11:18:51
Output data:
event: content_block_delta
[TRACE] 2025-01-29 11:18:51
Output data:
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"'ll add a brief comment at the top"}         }
[TRACE] 2025-01-29 11:18:51
Output data:
event: content_block_delta
[TRACE] 2025-01-29 11:18:51
Output data:
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" of the file explaining its purpose."}    }
[TRACE] 2025-01-29 11:18:51
Output data:
event: content_block_delta
[TRACE] 2025-01-29 11:18:51
Output data:
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"\n\nxml\n<tools"}    }
[TRACE] 2025-01-29 11:18:51
Output data:
event: content_block_delta
[TRACE] 2025-01-29 11:18:51
Output data:
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":">\n  <tool name=\"editor"} }
[TRACE] 2025-01-29 11:18:51
Output data:
event: content_block_delta
[TRACE] 2025-01-29 11:18:51
Output data:
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"\">\n    <action type"}            }
[TRACE] 2025-01-29 11:18:52
Output data:
event: content_block_delta
[TRACE] 2025-01-29 11:18:52
Output data:
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"=\"add\">\n      <line"}            }
[TRACE] 2025-01-29 11:18:52
Output data:
event: content_block_delta
[TRACE] 2025-01-29 11:18:52
Output data:
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":">1</line>\n      <code"}     }
[TRACE] 2025-01-29 11:18:52
Output data:
event: content_block_delta
[TRACE] 2025-01-29 11:18:52
Output data:
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"><![CDATA["}               }
[TRACE] 2025-01-29 11:18:52
Output data:
event: content_block_delta
[TRACE] 2025-01-29 11:18:52
Output data:
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"\"\"\"Builds version matrices for coh"}         }
[TRACE] 2025-01-29 11:18:52
Output data:
event: content_block_delta
[TRACE] 2025-01-29 11:18:52
Output data:
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"orts to visualize OTA update"}       }
[TRACE] 2025-01-29 11:18:52
Output data:
event: content_block_delta
[TRACE] 2025-01-29 11:18:52
Output data:
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" paths and device states.\"\"\"]]></code>"}               }
[TRACE] 2025-01-29 11:18:53
Output data:
event: content_block_delta
[TRACE] 2025-01-29 11:18:53
Output data:
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"\n      <buffer>1</buffer>\n    "}  }
[TRACE] 2025-01-29 11:18:53
Output data:
event: content_block_delta
[TRACE] 2025-01-29 11:18:53
Output data:
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"</action>\n  </tool>\n</"}        }
[TRACE] 2025-01-29 11:18:53
Output data:
event: content_block_delta
[TRACE] 2025-01-29 11:18:53
Output data:
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"tools>\n\n\nThis adds"}            }
[TRACE] 2025-01-29 11:18:53
Output data:
event: content_block_delta
[TRACE] 2025-01-29 11:18:53
Output data:
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" a concise docstring that"}           }
[TRACE] 2025-01-29 11:18:53
Output data:
event: content_block_delta
[TRACE] 2025-01-29 11:18:53
Output data:
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" explains the main purpose of this"}    }
[TRACE] 2025-01-29 11:18:54
Output data:
event: content_block_delta
[TRACE] 2025-01-29 11:18:54
Output data:
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" module - it handles building version matrices that"}    }
[TRACE] 2025-01-29 11:18:54
Output data:
event: content_block_delta
[TRACE] 2025-01-29 11:18:54
Output data:
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" show how devices can move between different software"}  }
[TRACE] 2025-01-29 11:18:54
Output data:
event: content_block_delta
[TRACE] 2025-01-29 11:18:54
Output data:
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" versions through OTA updates in"}             }
[TRACE] 2025-01-29 11:18:54
Output data:
event: content_block_delta
[TRACE] 2025-01-29 11:18:54
Output data:
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" a cohort."}   }
[TRACE] 2025-01-29 11:18:54
Output data:
event: content_block_delta
[TRACE] 2025-01-29 11:18:54
Output data:
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"\n\nSuggestions for next steps:"}       }
[TRACE] 2025-01-29 11:18:54
Output data:
event: content_block_delta
[TRACE] 2025-01-29 11:18:54
Output data:
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"\n1. Add more detailed "}      }
[TRACE] 2025-01-29 11:18:55
Output data:
event: content_block_delta
[TRACE] 2025-01-29 11:18:55
Output data:
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"docstrings to key classes"}   }
[TRACE] 2025-01-29 11:18:55
Output data:
event: content_block_delta
[TRACE] 2025-01-29 11:18:55
Output data:
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"\n2. Add type hints to function"}           }
[TRACE] 2025-01-29 11:18:55
Output data:
event: content_block_delta
[TRACE] 2025-01-29 11:18:55
Output data:
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" parameters\n3. Add return type hints to functions"}         }
[TRACE] 2025-01-29 11:18:55
Output data:
event: content_block_delta
[TRACE] 2025-01-29 11:18:55
Output data:
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":"\n4. Break up some of"}    }
[TRACE] 2025-01-29 11:18:55
Output data:
event: content_block_delta
[TRACE] 2025-01-29 11:18:55
Output data:
data: {"type":"content_block_delta","index":0,"delta":{"type":"text_delta","text":" the longer methods into smaller ones"}               }
[TRACE] 2025-01-29 11:18:55
Output data:
event: content_block_stop
[TRACE] 2025-01-29 11:18:55
Output data:
data: {"type":"content_block_stop","index":0           }
[TRACE] 2025-01-29 11:18:55
Output data:
event: message_delta
[TRACE] 2025-01-29 11:18:55
Output data:
data: {"type":"message_delta","delta":{"stop_reason":"end_turn","stop_sequence":null},"usage":{"output_tokens":200}     }
[TRACE] 2025-01-29 11:18:55
Output data:
event: message_stop
[TRACE] 2025-01-29 11:18:55
Output data:
data: {"type":"message_stop"          }

It stops there.

[fnune]

I've tried:

• Disabling all my plugins
• Removing all my configuration files one by one
• Replacing nvim-cmp with blink.cmp

Nothing fixes it, however on the minimal config it works. Maybe there's some cache file I can remove?

[fnune]

With the minimal config, the log file moves on to:

[DEBUG] 2025-01-29 11:38:44
Tool detected: { '<tools>\n  <tool name="editor">\n    <action type="add">\n      <line>1</line>\n      <code><![CDATA[-- CodeCompanion configuration for AI-assisted coding in Neovim]]></code>\n      <buffer>1</buffer>\n    </action>\n  </tool>\n</tools>' }
[TRACE] 2025-01-29 11:38:44
Trying to parse: <tools>
  <tool name="editor">
    <action type="add">
      <line>1</line>
      <code><![CDATA[-- CodeCompanion configuration for AI-assisted coding in Neovim]]></code>
      <buffer>1</buffer>
    </action>
  </tool>
</tools>
[TRACE] 2025-01-29 11:38:44
Parsed xml: {
  tools = {
    tool = {
      _attr = {
        name = "editor"
      },
      action = {
        _attr = {
          type = "add"
        },
        buffer = "1",
        code = "-- CodeCompanion configuration for AI-assisted coding in Neovim",
        line = "1"
      }
    }
  }
}
[DEBUG] 2025-01-29 11:38:44
Calling tool: strategies.chat.tools.editor
[DEBUG] 2025-01-29 11:38:44
Agent started
[DEBUG] 2025-01-29 11:38:44
Running cmd: function: 0x7fd5de27f220
[DEBUG] 2025-01-29 11:38:44
Editor tool request: {
  _attr = {
    type = "add"
  },
  buffer = "1",
  code = "-- CodeCompanion configuration for AI-assisted coding in Neovim",
  line = "1"
}
[TRACE] 2025-01-29 11:38:44
Using default diff
[TRACE] 2025-01-29 11:38:44
Using default diff
[TRACE] 2025-01-29 11:38:44
Adding code to buffer
[INFO] 2025-01-29 11:38:44
Chat request completed
[DEBUG] 2025-01-29 11:38:44
Agent Finished
[DEBUG] 2025-01-29 11:38:44
Tool editor finished
[DEBUG] 2025-01-29 11:38:44
Agent finished

[fnune]

Okay, I've figured it out. It's probably my plenary being pinned to version = "*" and that resulting in a version that's too old.

[fnune]

Here it is: fnune/home.nix@6fcacd9

Thanks for bearing with me, hope this helps someone else out there.

[fnune]

See #749 (reply in thread)

tl;dr: make sure plenary isn't pinned to some version, CodeCompanion needs a fairly fresh version of it.

[fnune]

@olimorris Perhaps it could be documented which features of plenary are needed, and when those features were introduced?

[olimorris]

This was noted in #344, added to the README and then seemingly lost when I moved to the docs site. I will add back and thanks for your diligence on this

[fnune]

No thank you!

[fnune]

It could perhaps even be checked at runtime. It seems like maybe a feature was missing and CodeCompanion was failing silently?