Context Window Manager #424
Description
Aims
- retry out of context window tool response failures with prompt shim to reduce context.
- track the context window growth over time and create a prompt shim and a tool which comes in at 80%. this tool condenses the prompt and starts a new agent.
- catch these errors and display them to a user better.
inspo
Discussed in #423
Originally posted by mherbertecifm May 13, 2025
I have an action job configured like this;
name: Shippie 🚢
on:
pull_request:
branches: [develop]
jobs:
run_code_review:
runs-on: ubuntu-latest
steps:
- uses: actions/checkout@v3
with:
fetch-depth: 0
- name: Shippie 🚢
uses: mattzcarey/[email protected]
with:
GITHUB_TOKEN: ${{ github.token }}
OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}
MODEL_STRING: "openai:gpt-4o"
I even have this configured to use v0.10.0, but it seems that shippie is pulling latest - that's a different issue, but it kinda leads into this breakage.
the config above has not changed. all PRs were being processed without error - until now.
here's the log snippet
...
...
installed [email protected] with binaries:
- shippie
2380 packages installed [19.92s]
INFO Using GitHub platform provider.
INFO Shippie collects anonymous usage data to help improve the product. You can opt out by setting --telemetry false when running Shippie.
INFO Running agentic review (max retries: 3)...
ERROR No MCP configuration found in .shippie/mcp.json or .cursor/mcp.json
ERROR Cannot start clients: MCP config not loaded
INFO Attempt 1/3...
ERROR Review failed with error:
AI_APICallError This model's maximum context length is 128000 tokens. However, your messages resulted in 784547 tokens (783388 in the messages, 1159 in the functions). Please reduce the length of the messages or functions., AI_APICallError, , https://api.openai.com/v1/chat/completions, [object Object], 400, [object Object], {
"error": {
"message": "This model's maximum context length is 128000 tokens. However, your messages resulted in 784547 tokens (783388 in the messages, 1159 in the functions). Please reduce the length of the messages or functions.",
"type": "invalid_request_error",
"param": "messages",
"code": "context_length_exceeded"
}
}, false, [object Object]
error stack:
...
...
this would imply that the MCP config is required?