mirror of https://github.com/langchain-ai/deepagents.git synced 2026-07-21 01:05:27 -04:00

Files

T

Mason Daugherty 3d1a9a9450 feat(infra): expand GitHub Action inputs (#4614 )

The Deep Agents Code GitHub Action now supports headless options for
model configuration, skills, startup commands, sandboxes, MCP,
interpreter tools, turn and task limits, rubric grading, stdin, quiet
output, buffered output, and JSON output.

---

The root GitHub Action exposed only a small subset of headless `dcode`
options, so workflows could not configure skills, sandboxes, MCP,
interpreter tools, rubric grading, model parameters, or output modes
without bypassing the Action.

This expands the Action input surface to match supported headless
`dcode` behavior while deliberately keeping interactive-only
`--auto-approve` out of workflows. Inputs are assembled as shell
argument arrays, typed values are validated before launch, stdin and
timeout handling preserve the agent's actual exit code, and memory
caching now includes SQLite WAL files. `ACTION.md` documents the common
inputs, outputs, memory behavior, and the need to treat agent output as
untrusted.

The Action tests compare forwarded flags with the CLI parser and execute
the real command-assembly shell against stubbed processes, covering
validation, timeout arithmetic, stdin behavior, and exit-code
propagation.

2026-07-09 18:06:03 -04:00

2.4 KiB

Raw Permalink Blame History

Deep Agents Code GitHub Action

This repository's root action.yml runs dcode non-interactively in GitHub Actions.

Example

name: dcode
on:
  workflow_dispatch:

permissions:
  contents: read

jobs:
  run:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@9c091bb21b7c1c1d1991bb908d89e4e9dddfe3e0 # v7.0.0
      - uses: langchain-ai/deepagents@main
        with:
          prompt: "Review this repository and summarize the highest-risk issues."
          model: "openai:gpt-5.5"
          openai_api_key: ${{ secrets.OPENAI_API_KEY }}
          shell_allow_list: "recommended,git,gh"
          max_turns: "8"
          quiet: "true"

For production workflows, pin langchain-ai/deepagents to a reviewed commit SHA instead of main.

Common inputs

prompt: Task text passed to dcode.
model: Model as provider:model (e.g. openai:gpt-5.5) or a bare name (claude-*, gpt-*, gemini-*) with the provider auto-detected.
*_api_key: Provider API keys (openai_api_key, anthropic_api_key, google_api_key).
shell_allow_list: Commands allowed for headless shell execution, such as recommended,git,gh.
max_turns: Maximum agentic turns before stopping.
task_timeout: dcode --timeout in seconds. The action-level timeout input is in minutes.
quiet: Clean stdout output for piping.
json: Emit machine-readable dcode JSON output.

Advanced inputs

The action also forwards headless dcode options for skills, startup commands, sandboxes, MCP config, interpreter tools, rubric grading, stdin, and model/profile overrides. Interactive-only options such as --auto-approve are intentionally not exposed; use shell_allow_list to permit shell commands in workflows. See action.yml for the full input list and defaults.

Outputs

response: The agent's full output (stdout and stderr combined). This is raw, unfiltered agent output — do not assume it is free of secrets or safe to echo into other services.
exit_code: The agent's exit code (also 124 on task timeout).
cache_hit: Whether agent memory was restored from cache (empty when enable_memory is false).

Memory

Persistent memory is enabled by default through actions/cache. Use:

enable_memory: "false" to disable cache restore/save.
memory_scope: pr, branch, or repo to control cache sharing.
agent_name to separate memory between multiple action identities.

2.4 KiB Raw Permalink Blame History