GEPA
[gepa] + [proposer] in gepa.toml. Shared TOML: Optimizers. Container routes: Containers.
code
[proposer]backend = "codex_app_server"execution_mode = "local_process"model = "gpt-5.4-nano"auth_mode = "api_key"api_key_env = "OPENAI_API_KEY"copy_host_auth = falsesandbox_mode = "workspace-write"approval_policy = "never"timeout_seconds = 900reasoning_effort = "medium"| Field | Type | Notes |
|---|---|---|
| backend | string | must be codex_app_server |
| execution_mode | string | local_process |
| model | string | |
| auth_mode | string | api_key |
| api_key_env | string | env var for Codex key |
| copy_host_auth | bool | false for API-key cookbooks |
| sandbox_mode | string | workspace-write |
| approval_policy | string | never |
| timeout_seconds | integer | |
| reasoning_effort | string | optional |
code
[gepa]max_generations = 3proposals_per_generation = 6minibatch_size = 40max_total_rollouts = 4096max_train_rollouts = 520max_heldout_rollouts = 220rollout_submission_mode = "sync"max_cost_usd = 0.0| Field | Type | Notes |
|---|---|---|
| max_generations | integer | |
| proposals_per_generation | integer | |
| minibatch_size | integer | |
| minibatch_accept_margin | float | default 0 |
| max_total_rollouts | integer | |
| max_train_rollouts | integer | optional |
| max_heldout_rollouts | integer | optional |
| rollout_submission_mode | sync | async | |
| rollout_poll_interval_ms | integer | async |
| rollout_async_timeout_seconds | integer | async |
| rollout_chunk_size | integer | optional |
| frontier_type | string | |
| acceptance_criterion | string | |
| max_cost_usd | float | 0 = off |
| max_time_seconds | integer | optional stop |
| no_improvement_generations | integer | optional stop |
code
[gepa.pipeline]mode = "sync_serial"[gepa.pipeline.workers]rollout = 20max_in_flight_candidates = 6[gepa.pipeline.adaptive_rollout_concurrency]enabled = trueinitial = 50min = 1max = 120- gepa.pipeline.mode
- sync_serial | async_pipelined
- gepa.pipeline.workers.rollout
- integer
- gepa.pipeline.workers.max_in_flight_candidates
- integer
- gepa.pipeline.adaptive_rollout_concurrency.*
- enabled, initial, min, max, increase_step, decrease_step, increase_after_successes
| Env | Overrides |
|---|---|
| SYNTH_OPTIMIZERS_GEPA_PIPELINE_MODE | gepa.pipeline.mode |
| SYNTH_OPTIMIZERS_GEPA_ROLLOUT_SUBMISSION_MODE | gepa.rollout_submission_mode |
| SYNTH_OPTIMIZERS_GEPA_ROLLOUT_CHUNK_SIZE | gepa.rollout_chunk_size |
| SYNTH_OPTIMIZERS_GEPA_ROLLOUT_HTTP_RETRIES | rollout HTTP retries |
| OPENAI_API_KEY | proposer (when proposer.api_key_env=OPENAI_API_KEY) |
code
bash run_fresh_gepa.sh --profile smokesynth-optimizers gepa run --config gepa.toml