Reference
Inputs & Outputs
The config-driven input/output contract
trajgen is configured entirely through config.yaml. It uses config.yaml rather than a separate inputs.yaml because the file describes a full Harbor run profile, not just upstream inputs. Treat config.yaml as the source of truth before launching a job.
Inputs
| Name | Description | Source |
|---|---|---|
repositories | Git remote/ref/path/read-only policy for harbor and swe_data_process | external |
environment | Harbor uv env, LiteLLM venv, and swe_data_process uv env | external |
llm_api | Raw upstream model API used by the per-job LiteLLM proxy | external |
litellm_proxy | LiteLLM config template, port, and master key | external |
task_source | SWE task source (wired from swegen) | dependency |
harbor_job | jobs dir, concurrency, retries, timeout multiplier, task cap | external |
agent | Harbor agent scaffold, version, runtime image, sampling controls | external |
sft_conversion | Optional post-job conversion step | external |
HARBOR_EXCLUDE_TASKS | Space-separated task IDs Harbor must skip | derived |
The only external value you normally fill is runtime_info.input.llm_api; the task source is wired from swegen via meta_info.dependencies.
Active runtime values
Excerpted from config.yaml:
llm_api:
api_key: dummy-key
api_base_url: "http://llm10.jierungogogo.com/v1"
model: "openai/MiniMax-M2.7"
protocols: [openai_compatible, anthropic_compatible]
served_via: per_job_litellm_proxy
litellm_proxy:
config_template: scripts/serve_llm/litellm_config.example.yaml
port: 4001
master_key: dummy-key-cf
task_source:
provider: local
dataset_name: ../swegen/artifacts/swe_tasks/py-cc
harbor_job:
jobs_dir: artifacts/jobs
n_concurrent: 8
n_tasks: 16
max_retries: 2
timeout_multiplier: 2
agent:
name: custom-claude-code
version: 2.1.118
runtime_image: docker.io/jierun/c-cc-2.1.118:v0.1
max_turns: 80
temperature: 0.7
sft_conversion:
enabled: false
scaffold: auto
out_dir: artifacts/sft_dataOutputs
trajgen declares its handoff contract in config.yaml → runtime_info.output:
| Output | Path | Format | Consumer |
|---|---|---|---|
raw_trajectories_dir | artifacts/jobs/ | artifacts/jobs/<job>/<task>/agent/litellm-trajectory.jsonl | sft |
sft_data_dir | artifacts/sft_data/ | artifacts/sft_data/<job>/lf.json (LLaMA-Factory) | sft |
See Results & Artifacts for the full layout, and Config Variants for running experiments without disturbing the active profile.