Skip to content

Bound OpenCode review cadence and scheduler backlog #383

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Closed
seonghobae wants to merge 4 commits into from
Closed

Bound OpenCode review cadence and scheduler backlog #383

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
2267929
Bound OpenCode review cadence
seonghobae Jul 9, 2026
1805b1b
Tune OpenCode review cadence for deep reviews
seonghobae Jul 9, 2026
95112c6
Clarify OpenCode DeepSeek session default finding
seonghobae Jul 9, 2026
7916854
ci: refresh opencode cadence checks
seonghobae Jul 9, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
38 changes: 15 additions & 23 deletions .github/workflows/opencode-review.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -2335,35 +2335,32 @@ jobs:
- name: Run OpenCode PR Review model pool
id: opencode_review_model_pool
if: needs.coverage-evidence.result == 'success'
timeout-minutes: 350
timeout-minutes: 330
continue-on-error: true
Comment on lines 2335 to +2339
env:
STRIX_GITHUB_MODELS_TOKEN: ${{ secrets.STRIX_GITHUB_MODELS_TOKEN || github.token }}
GITHUB_TOKEN: ${{ secrets.STRIX_GITHUB_MODELS_TOKEN || github.token }}
USE_GITHUB_TOKEN: "true"
SHARE: "false"
NPM_CONFIG_IGNORE_SCRIPTS: "true"
NO_COLOR: "1"
# Ordered contract-reliability first, then quota, with the rate-starved
# flagships last. The mini reasoning models (o4-mini, o3-mini, gpt-5-
# mini/nano/chat) reliably emit the strict review contract — every
# required label and only source-backed findings — so they lead. The
# high-quota non-reasoning models (deepseek-v3, mistral, llama-4) emit
# bare or hallucinated reviews the publish/approve gates reject, so
# they are fallbacks only. gpt-5/o3 ("Reasoning" tier, 8-12 req/day)
# stay last: first-placing them stalled every review until timeout
# because a rate-limited/hung flagship never fell back.
OPENCODE_MODEL_CANDIDATES: "github-models/openai/o4-mini github-models/openai/o3-mini github-models/openai/gpt-5-mini github-models/openai/gpt-5-nano github-models/openai/gpt-5-chat github-models/deepseek/deepseek-r1-0528 github-models/deepseek/deepseek-r1 github-models/deepseek/deepseek-v3-0324 github-models/mistral-ai/mistral-medium-2505 github-models/meta/llama-4-maverick-17b-128e-instruct-fp8 github-models/meta/llama-4-scout-17b-16e-instruct github-models/openai/o3 github-models/openai/gpt-5"
# Ordered by observed successful review completion first, then
# contract-reliability and quota. DeepSeek V3 currently has the best
# success rate in the org queue, so it leads with enough time to finish
# a deep review. Compact reasoning models remain early fallbacks. The
# rate-starved flagships stay last because first-placing them stalled
# previous reviews until timeout instead of reaching healthier models.
OPENCODE_MODEL_CANDIDATES: "github-models/deepseek/deepseek-v3-0324 github-models/openai/o4-mini github-models/openai/o3-mini github-models/openai/gpt-5-mini github-models/openai/gpt-5-nano github-models/openai/gpt-5-chat github-models/deepseek/deepseek-r1-0528 github-models/deepseek/deepseek-r1 github-models/mistral-ai/mistral-medium-2505 github-models/meta/llama-4-maverick-17b-128e-instruct-fp8 github-models/meta/llama-4-scout-17b-16e-instruct github-models/openai/o3 github-models/openai/gpt-5"
# One attempt per model, then fall through to the next model. Retrying
# the SAME model 5x let a rate-limited/hung leader consume the whole
# step, so the pool never reached a healthy fallback model.
OPENCODE_MODEL_ATTEMPTS: "1"
# 90 min per model — generous for a deep tool-using review, but bounded
# so a rate-limited model yields to the next one instead of eating the
# 350-min step. (20400s = 340min gave one model the entire budget with
# no fallback; 600s was too short for a proper review.)
# 90 min per model gives deep reviews enough room. The pool budget,
# not a short per-model cutoff, prevents stale provider calls from
# consuming the whole 360-min job before the publish gate logs why.
OPENCODE_RUN_TIMEOUT_SECONDS: "5400"
OPENCODE_EXPORT_TIMEOUT_SECONDS: "120"
OPENCODE_TOTAL_RETRY_BUDGET_SECONDS: "0"
OPENCODE_TOTAL_RETRY_BUDGET_SECONDS: "18000"
OPENCODE_BACKOFF_INITIAL_SECONDS: "30"
OPENCODE_BACKOFF_MAX_SECONDS: "30"
OPENCODE_FIRST_ATTEMPT_AGENT: ci-review
Expand Down Expand Up @@ -2720,12 +2717,7 @@ jobs:
fi

- name: Approve PR if OpenCode review passed
if: >-
always()
&& (
needs.coverage-evidence.result != 'success'
|| steps.opencode_review_model_pool.outcome == 'success'
)
if: always()
Comment on lines 2719 to +2720
timeout-minutes: 75
env:
GH_TOKEN: ${{ secrets.OPENCODE_APPROVE_TOKEN || steps.opencode_app_token.outputs.token || github.token }}
Expand Down Expand Up @@ -3611,7 +3603,7 @@ jobs:
"scripts/ci/test_strix_quick_gate.sh"
emit_known_missing_string_finding \
"MODEL: github-models/deepseek/deepseek-r1-0528" \
"OpenCode review must start with DeepSeek R1" \
"OpenCode review config must keep the DeepSeek R1 session default" \
".github/workflows/opencode-review.yml" \
"scripts/ci/test_strix_quick_gate.sh"

Expand Down
4 changes: 2 additions & 2 deletions .github/workflows/pr-review-merge-scheduler.yml
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -33,7 +33,7 @@ on:
review_dispatch_limit:
description: Maximum OpenCode/Strix review dispatch actions per scheduler run
required: false
default: "1"
default: "-1"
type: string
enable_auto_merge:
description: Enable auto-merge for current-head approved PRs
Expand Down Expand Up @@ -95,7 +95,7 @@ on:
review_dispatch_limit:
description: Maximum OpenCode/Strix review dispatch actions per scheduler run
required: false
default: "1"
default: "-1"
enable_auto_merge:
description: Enable auto-merge for current-head approved PRs
required: false
Expand Down
18 changes: 9 additions & 9 deletions scripts/ci/test_strix_quick_gate.sh
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -521,16 +521,16 @@ assert_opencode_review_uses_codegraph_and_gpt5_fallback() {
assert_file_contains "$REPO_ROOT/scripts/ci/run_opencode_review_model_pool.sh" "tokens_limit_reached" "opencode review detects provider context-window overflow"
assert_file_contains "$REPO_ROOT/scripts/ci/run_opencode_review_model_pool.sh" "skipping remaining attempts for this model" "opencode review skips same-model retries after context-window overflow"
assert_file_contains "$workflow_file" 'timeout-minutes: 360' "opencode review target uses the maximum GitHub-hosted runner timeout"
assert_file_contains "$workflow_file" 'timeout-minutes: 285' "opencode model pool keeps retrying for most of the job budget while leaving approval headroom"
assert_file_contains "$workflow_file" 'OPENCODE_RUN_TIMEOUT_SECONDS: "600"' "opencode primary review has a bounded per-model timeout before trying fallback models"
assert_file_contains "$workflow_file" 'OPENCODE_TOTAL_RETRY_BUDGET_SECONDS: "0"' "opencode model pool has no script-level retry budget"
assert_file_contains "$workflow_file" 'timeout-minutes: 330' "opencode model pool keeps retrying with bounded queue impact while leaving approval headroom"
assert_file_contains "$workflow_file" 'OPENCODE_RUN_TIMEOUT_SECONDS: "5400"' "opencode primary review has a deep per-model timeout before trying fallback models"
assert_file_contains "$workflow_file" 'OPENCODE_TOTAL_RETRY_BUDGET_SECONDS: "18000"' "opencode model pool exits before the job timeout so the approval gate can publish a reason"
assert_file_contains "$workflow_file" "needs.coverage-evidence.result == 'success'" "opencode model pool only runs after coverage evidence passed"
assert_file_contains "$workflow_file" "id: opencode_review_model_pool" "opencode DeepSeek V3 fallback still runs after a primary model timeout or step failure when coverage evidence passed"
assert_file_contains "$workflow_file" "always()" "opencode fallback chain uses always() so failed model steps cannot skip every fallback"
assert_file_contains "$workflow_file" 'OPENCODE_MODEL_ATTEMPTS: "1"' "opencode fallback tries the catalog promptly instead of spending the entire review on one model"
assert_file_contains "$workflow_file" "Run OpenCode PR Review model pool" "opencode review includes a broad catalog fallback pool"
assert_file_contains "$workflow_file" "steps.opencode_review_model_pool.outcome == 'success'" "opencode model step must succeed before review publication"
assert_file_contains "$workflow_file" "github-models/openai/o4-mini github-models/openai/o3-mini github-models/openai/gpt-5-mini github-models/openai/gpt-5-chat github-models/openai/o3 github-models/mistral-ai/mistral-medium-2505 github-models/openai/gpt-5-nano github-models/deepseek/deepseek-r1-0528 github-models/deepseek/deepseek-r1 github-models/deepseek/deepseek-v3-0324 github-models/meta/llama-4-maverick-17b-128e-instruct-fp8 github-models/meta/llama-4-scout-17b-16e-instruct" "opencode review tries high-effort reasoning fallbacks before broader catalog models"
assert_file_not_contains "$workflow_file" "steps.opencode_review_model_pool.outcome == 'success'" "opencode approval gate still runs after model pool failure to publish a reason"
assert_file_contains "$workflow_file" "github-models/deepseek/deepseek-v3-0324 github-models/openai/o4-mini github-models/openai/o3-mini github-models/openai/gpt-5-mini" "opencode review starts with the observed high-success DeepSeek V3 path before compact reasoning fallbacks"
assert_file_contains "$workflow_file" "The publish gate re-runs source-backed validation against PR-head data" "opencode review publish gate validates model output against the PR-head worktree"
assert_file_contains "$workflow_file" '"openai/o3"' "opencode config declares OpenAI o3 fallback"
assert_file_contains "$workflow_file" '"openai/o4-mini"' "opencode config declares OpenAI o4-mini fallback"
Expand Down Expand Up @@ -632,11 +632,11 @@ assert_opencode_review_uses_codegraph_and_gpt5_fallback() {
assert_file_not_contains "$workflow_file" "no model produced a valid review control block" "opencode model-failure path no longer documents a final exhausted state"
assert_file_contains "$workflow_file" 'OPENCODE_MODEL_ATTEMPTS: "1"' "opencode primary and fallback paths avoid multi-attempt stalls on one model"
assert_file_contains "$workflow_file" 'OPENCODE_MODEL_ATTEMPTS: "1"' "opencode catalog fallback tries each model once before moving on"
assert_file_contains "$workflow_file" 'OPENCODE_RUN_TIMEOUT_SECONDS: "600"' "opencode catalog fallback has a bounded model review timeout before step timeout"
assert_file_contains "$workflow_file" 'OPENCODE_RUN_TIMEOUT_SECONDS: "5400"' "opencode catalog fallback has a bounded deep-review timeout before step timeout"
assert_file_contains "$REPO_ROOT/scripts/ci/run_opencode_review_model_pool.sh" "OpenCode %s attempt %s/%s failed" "opencode catalog fallback records per-model retry failures"
assert_file_contains "$REPO_ROOT/scripts/ci/run_opencode_review_model_pool.sh" "exponential backoff" "opencode model retry paths use exponential backoff instead of fixed sleeps"
assert_file_contains "$workflow_file" "github-models/openai/o4-mini github-models/openai/o3-mini" "opencode review tries compact OpenAI reasoning model fallbacks early"
assert_file_contains "$workflow_file" "github-models/openai/gpt-5-chat github-models/openai/o3 github-models/mistral-ai" "opencode review keeps full OpenAI and non-OpenAI catalog fallbacks after compact reasoning attempts"
assert_file_contains "$workflow_file" "github-models/deepseek/deepseek-v3-0324 github-models/openai/o4-mini github-models/openai/o3-mini" "opencode review tries DeepSeek V3 before compact OpenAI reasoning model fallbacks"
assert_file_contains "$workflow_file" "github-models/openai/gpt-5-chat github-models/deepseek/deepseek-r1-0528 github-models/deepseek/deepseek-r1" "opencode review keeps full reasoning-capable catalog fallbacks after compact attempts"
assert_file_contains "$workflow_file" "coverage-evidence:" "opencode workflow measures coverage before review"
assert_file_contains "$workflow_file" "github.event_name == 'workflow_dispatch' || github.event_name == 'pull_request_target'" "manual and required OpenCode reviews measure coverage instead of approving skipped coverage evidence"
assert_file_contains "$workflow_file" "Exchange OpenCode app token for target repository coverage reads" "coverage evidence can read private target repositories through the OpenCode app token"
Expand Down Expand Up @@ -824,7 +824,7 @@ assert_opencode_review_uses_codegraph_and_gpt5_fallback() {
assert_file_contains "$workflow_file" 'GH_TOKEN: ${{ secrets.OPENCODE_APPROVE_TOKEN || steps.review_read_app_token.outputs.token || github.token }}' "opencode manual dispatch uses the cross-repo approval token for target PR evidence lookups with app-token fallback"
assert_file_contains "$workflow_file" 'repos/${GH_REPOSITORY}' "opencode review workflow uses env-backed repository context in shell commands"
assert_file_contains "$workflow_file" "Run OpenCode PR Review model pool" "opencode review starts the central model pool"
assert_file_contains "$workflow_file" "github-models/openai/o4-mini github-models/openai/o3-mini github-models/openai/gpt-5-mini" "opencode review starts with faster reasoning-capable GitHub Models"
assert_file_contains "$workflow_file" "github-models/deepseek/deepseek-v3-0324 github-models/openai/o4-mini github-models/openai/o3-mini" "opencode review starts with high-success DeepSeek V3 before faster reasoning-capable GitHub Models"
assert_file_contains "$workflow_file" "github-models/deepseek/deepseek-r1-0528" "opencode review keeps a reachable DeepSeek R1 reasoning fallback model"
assert_file_contains "$workflow_file" "github-models/deepseek/deepseek-v3-0324" "opencode review has a reachable DeepSeek V3 fallback model"
assert_file_contains "$workflow_file" "github-models/openai/gpt-5" "opencode review still has a bounded GPT-5 fallback model"
Expand Down
13 changes: 7 additions & 6 deletions tests/test_opencode_agent_contract.py
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -79,9 +79,9 @@ def test_opencode_model_pool_sets_high_effort_for_capable_candidates():
assert candidate_models
assert set(candidate_models).issubset(set(models))
assert candidate_models[:3] == [
"deepseek/deepseek-v3-0324",
"openai/o4-mini",
"openai/o3-mini",
"openai/gpt-5-mini",
]
assert {
"openai/gpt-5-chat",
Expand Down Expand Up @@ -280,20 +280,21 @@ def test_workflow_provisions_sandbox_tool_and_reviewer_agent():
assert '"## Review outcome"' in workflow
assert '"## Check outcome"' not in workflow
assert "publish REQUEST_CHANGES when coverage-evidence blocker states" in workflow
assert re.search(r"opencode-review-target:[\s\S]{0,240}timeout-minutes: 360", workflow)
assert re.search(r"opencode-review-target:[\s\S]{0,400}timeout-minutes: 360", workflow)
assert 'timeout-minutes: 75' in workflow
assert re.search(r"Run OpenCode PR Review model pool[\s\S]{0,240}timeout-minutes: 350", workflow)
assert re.search(r"Run OpenCode PR Review model pool[\s\S]{0,260}timeout-minutes: 330", workflow)
assert re.search(r"Run OpenCode PR Review model pool[\s\S]{0,320}continue-on-error: true", workflow)
assert 'APPROVAL_CHECK_WAIT_ATTEMPTS: "81"' in workflow
assert 'APPROVAL_CHECK_WAIT_SLEEP_SECONDS: "30"' in workflow
assert (
'OPENCODE_MODEL_CANDIDATES: "github-models/openai/o4-mini '
'OPENCODE_MODEL_CANDIDATES: "github-models/deepseek/deepseek-v3-0324 '
"github-models/openai/o4-mini "
"github-models/openai/o3-mini "
"github-models/openai/gpt-5-mini "
"github-models/openai/gpt-5-nano "
'github-models/openai/gpt-5-chat '
"github-models/deepseek/deepseek-r1-0528 "
"github-models/deepseek/deepseek-r1 "
"github-models/deepseek/deepseek-v3-0324 "
"github-models/mistral-ai/mistral-medium-2505 "
"github-models/meta/llama-4-maverick-17b-128e-instruct-fp8 "
"github-models/meta/llama-4-scout-17b-16e-instruct "
Expand All @@ -303,7 +304,7 @@ def test_workflow_provisions_sandbox_tool_and_reviewer_agent():
assert 'OPENCODE_MODEL_ATTEMPTS: "1"' in workflow
assert 'OPENCODE_RUN_TIMEOUT_SECONDS: "5400"' in workflow
assert 'OPENCODE_EXPORT_TIMEOUT_SECONDS: "120"' in workflow
assert 'OPENCODE_TOTAL_RETRY_BUDGET_SECONDS: "0"' in workflow
assert 'OPENCODE_TOTAL_RETRY_BUDGET_SECONDS: "18000"' in workflow
assert 'OPENCODE_BACKOFF_MAX_SECONDS: "30"' in workflow
assert "while :" in model_pool_runner
assert "OpenCode model pool has no configured model candidates." in model_pool_runner
Expand Down
Loading