feat(stepfun): add step-3.7-flash model by lit26 · Pull Request #88082 · openclaw/openclaw

lit26 · 2026-05-29T17:15:38Z

What Problem This Solves

StepFun released step-3.7-flash, but OpenClaw's direct StepFun catalogs exposed only the Step 3.5 family. Users could not select the new model through either the Standard stepfun provider or the subscription-backed stepfun-plan provider.

Why This Change Was Made

Add step-3.7-flash to both official static catalogs while preserving the existing Step 3.5 defaults.
Match StepFun's published 256K context, Standard pricing, image input, tool-calling, and low/medium/high reasoning-effort contract.
Declare the actual Chat Completions wire behavior: system messages, max_tokens, no OpenAI-only store, stream_options, developer-role, or strict-tool fields.
Map OpenClaw's unsupported off, minimal, xhigh, and max levels to StepFun's nearest supported effort.

User Impact

Users can select stepfun/step-3.7-flash or stepfun-plan/step-3.7-flash. Existing installations remain on their current Step 3.5 default until the user explicitly switches. OpenClaw advertises text and image input; StepFun's native video input remains outside OpenClaw's current model-input schema.

Evidence

Official model contract: https://platform.stepfun.ai/docs/en/guides/models/step-3.7-flash
Official Step Plan support and endpoint: https://platform.stepfun.ai/docs/en/step-plan/integrations/reasoning-api
Exact head: 0ad970c66111a7974954c85e6f54afcf7248dd6e.
Sanitized public-AWS Crabbox run run_564f467801fc (lease cbx_99addb986ee6; no secrets, instance role, tailnet, or LAN access):
- pnpm exec oxfmt --check docs/providers/stepfun.md extensions/stepfun/index.test.ts extensions/stepfun/openclaw.plugin.json src/agents/models-config.providers.stepfun.test.ts
- pnpm test extensions/stepfun/index.test.ts src/agents/models-config.providers.stepfun.test.ts
- Result: formatting clean; 6 tests passed across 2 Vitest shards.
Fresh Codex autoreview: clean, no accepted/actionable findings.
Direct inference was not run because no approved StepFun credential was available; API behavior is covered by the emitted-payload regression test and official contract above.

clawsweeper · 2026-05-29T17:17:11Z

Codex review: needs real behavior proof before merge. Reviewed July 9, 2026, 2:13 AM ET / 06:13 UTC.

Summary
The PR adds step-3.7-flash metadata to both bundled StepFun provider catalogs, updates StepFun docs, and adds focused provider/request-payload regression coverage while preserving Step 3.5 defaults.

PR surface: Source +80, Tests +93, Docs +65. Total +238 across 4 files.

Reproducibility: not applicable. as a bug reproduction; this is an additive provider model feature. Source inspection shows current main lacks step-3.7-flash, while the PR head adds it to both StepFun catalogs.

Review metrics: 1 noteworthy metric.

StepFun catalog defaults: 2 model rows added, 0 default refs changed. The change is additive for model selection and does not silently move existing StepFun users off the Step 3.5 defaults.

Merge readiness
Overall: 🦪 silver shellfish
Proof: 🦪 silver shellfish
Patch quality: 🦞 diamond lobster
Result: blocked until real behavior proof from a real setup is added.

Overall follows the weaker of proof and patch quality, so missing proof can cap an otherwise strong patch.

Rank-up moves:

[P1] Add redacted terminal output showing openclaw models list --provider stepfun and openclaw models list --provider stepfun-plan after the change, or a live step-3.7-flash inference transcript if credentials are available.
Wait for the remaining exact-head CI checks to complete successfully.

Proof guidance:

[P1] Needs real behavior proof before merge: The PR body and maintainer comment provide sanitized Crabbox formatting/test output and emitted-payload regression coverage, but no real OpenClaw catalog output or live StepFun inference proof for the new model. After adding proof, update the PR body; ClawSweeper should re-review automatically. If it does not, the PR author or someone with repository write access can comment @clawsweeper re-review.

Risk before merge

[P1] The current proof attached to the PR is targeted formatting and Vitest request-payload validation, not a real OpenClaw model-list run or redacted live StepFun inference transcript for the new model.
[P1] The current head is mergeable but behind and still had 6 checks in progress at review time, so exact-head CI needs to finish before merge.

Maintainer options:

Decide the mitigation before merge
Land the additive StepFun model metadata/docs/tests after exact-head CI is green and maintainers either receive lightweight real OpenClaw catalog/inference proof or explicitly accept the request-payload test proof for this provider-catalog update.
Pause or close
Do not merge this PR until maintainers decide whether the risk is worth taking.

Next step before merge

[P1] No automated code repair is indicated; the remaining action is maintainer proof acceptance plus exact-head CI/merge handling.

Maintainer decision needed

Question: Should this PR wait for real OpenClaw catalog or live StepFun proof, or may maintainers accept the current request-payload regression tests plus official StepFun docs for this catalog-only update?
Rationale: Repository policy treats external API/provider behavior as dependency-backed and external PRs normally need real behavior proof; this PR has strong metadata/tests and maintainer fixups, but no attached live catalog or inference output for step-3.7-flash.
Likely owner: steipete — This account is assigned to the PR and authored the recent PR-head compatibility/test fixups that made the change land-ready aside from proof and CI acceptance.
Options:
- Request Lightweight Real Proof (recommended): Ask for redacted terminal output showing openclaw models list for both StepFun providers, or a live step-3.7-flash inference transcript if credentials are available, before merge.
- Accept Current Proof: A maintainer can explicitly accept official StepFun docs plus emitted-payload tests as sufficient for this additive catalog update and merge once CI is green.
- Pause For Provider Owner Review: Hold the PR until a StepFun/provider owner confirms the model metadata and proof bar.

Security
Cleared: The diff changes bundled provider metadata, docs, and tests only; it adds no dependencies, workflows, permissions, credential handling, package scripts, or new code-execution path.

Review details

Best possible solution:

Land the additive StepFun model metadata/docs/tests after exact-head CI is green and maintainers either receive lightweight real OpenClaw catalog/inference proof or explicitly accept the request-payload test proof for this provider-catalog update.

Do we have a high-confidence way to reproduce the issue?

Not applicable as a bug reproduction; this is an additive provider model feature. Source inspection shows current main lacks step-3.7-flash, while the PR head adds it to both StepFun catalogs.

Is this the best way to solve the issue?

Yes for the code shape: the bundled StepFun plugin manifest, provider docs, and focused provider/request-payload tests are the right owner boundary. The remaining uncertainty is proof sufficiency, not a better implementation location.

AGENTS.md: found and applied where relevant.

Codex review notes: model internal, reasoning high; reviewed against a780541ea6c5.

Label changes

Label changes:

add rating: 🦪 silver shellfish: Overall readiness is 🦪 silver shellfish; proof is 🦪 silver shellfish and patch quality is 🦞 diamond lobster. Replaced prior rating: 🐚 platinum hermit.
remove proof: sufficient: Current real behavior proof status is mock_only, not sufficient.
remove rating: 🐚 platinum hermit: Current PR rating is rating: 🦪 silver shellfish, so this older rating label is no longer current.

Label justifications:

P2: This is a bounded provider-catalog feature with limited blast radius and no confirmed runtime regression in the patch.
rating: 🦪 silver shellfish: Overall readiness is 🦪 silver shellfish; proof is 🦪 silver shellfish and patch quality is 🦞 diamond lobster. Replaced prior rating: 🐚 platinum hermit.
status: 👀 ready for maintainer look: ClawSweeper has no concrete contributor-facing blocker left for this PR. Needs real behavior proof before merge: The PR body and maintainer comment provide sanitized Crabbox formatting/test output and emitted-payload regression coverage, but no real OpenClaw catalog output or live StepFun inference proof for the new model. After adding proof, update the PR body; ClawSweeper should re-review automatically. If it does not, the PR author or someone with repository write access can comment @clawsweeper re-review.

Evidence reviewed

PR surface:

Source +80, Tests +93, Docs +65. Total +238 across 4 files.

View PR surface stats

Area	Files	Added	Removed	Net
Source	1	80	0	+80
Tests	2	95	2	+93
Docs	1	73	8	+65
Config	0	0	0	0
Generated	0	0	0	0
Other	0	0	0	0
Total	4	248	10	+238

What I checked:

Repository policy read: Root and scoped AGENTS.md guidance for provider plugins, docs, and agent tests was read and applied to the provider-boundary, changelog, and dependency-contract review. (AGENTS.md:1, a780541ea6c5)
Current main lacks the requested model: Current main's StepFun manifest lists only the existing step-3.5-flash standard row and Step Plan 3.5 rows; step-3.7-flash is absent. (extensions/stepfun/openclaw.plugin.json:29, a780541ea6c5)
PR head adds Step 3.7 rows additively: The PR head adds step-3.7-flash to both stepfun and stepfun-plan with text/image input, 262144 context/max tokens, StepFun pricing for standard, zero plan cost matching existing plan rows, and compatibility flags for the Chat Completions payload. (extensions/stepfun/openclaw.plugin.json:29, 0ad970c66111)
Existing defaults are preserved: The PR head keeps STEPFUN_DEFAULT_MODEL_ID as step-3.5-flash, so existing StepFun onboarding/default model selection is not silently changed. (extensions/stepfun/provider-catalog.ts:14, 0ad970c66111)
Focused request-payload coverage exists: The PR adds tests that assert the new model metadata, unchanged defaults, reasoning_effort, max_tokens, omitted OpenAI-only fields, and system-message behavior for the StepFun OpenAI-compatible transport path. (extensions/stepfun/index.test.ts:39, 0ad970c66111)
Runtime consumes the compat metadata: getCompat reads model compat flags and the Chat Completions parameter builder chooses max_tokens and emits reasoning_effort only when the model supports it, matching the PR's metadata/test boundary. (src/agents/openai-transport-stream.ts:3633, a780541ea6c5)

Likely related people:

hengm3467: Authored the merged StepFun provider work that added the manifest, catalog builder, docs, and provider catalog tests extended by this PR. (role: introduced StepFun provider surface; confidence: high; commits: 52d8dc5b56c1, 326d9427197e; files: extensions/stepfun/openclaw.plugin.json, extensions/stepfun/provider-catalog.ts, docs/providers/stepfun.md)
odysseus0: GitHub metadata for the merged StepFun provider PR identifies this account as the merger of the original StepFun provider implementation. (role: original provider PR merger; confidence: high; commits: 326d9427197e; files: extensions/stepfun/openclaw.plugin.json, extensions/stepfun/provider-catalog.ts, docs/providers/stepfun.md)
steipete: This PR is assigned to this account, and this account authored the recent PR-head compatibility and test fixups that made the StepFun 3.7 payload behavior match the documented API contract. (role: current reviewer and recent adjacent contributor; confidence: high; commits: 670215315b77, 975b68cfdb92, 6d7386fcf2d6; files: extensions/stepfun/openclaw.plugin.json, extensions/stepfun/index.test.ts, docs/providers/stepfun.md)
Vincent Koc: Recent history includes StepFun provider docs rewrites and release-adjacent commits touching the same docs/manifest surfaces, making this a useful routing signal for docs/provider review. (role: recent provider-docs contributor; confidence: medium; commits: 81d32c05f431, e085fa1a3ffd; files: docs/providers/stepfun.md, extensions/stepfun/openclaw.plugin.json)

What the crustacean ranks mean

🦀 challenger crab: rare, exceptional readiness with strong proof, clean implementation, and convincing validation.
🦞 diamond lobster: very strong readiness with only minor maintainer review expected.
🐚 platinum hermit: good normal PR, likely mergeable with ordinary maintainer review.
🦐 gold shrimp: useful signal, but proof or patch confidence is still limited.
🦪 silver shellfish: thin signal; proof, validation, or implementation needs work.
🧂 unranked krab: not merge-ready because proof is missing/unusable or there are serious correctness or safety concerns.
🌊 off-meta tidepool: rating does not apply to this item.

Shiny media proof means a screenshot, video, or linked artifact directly shows the changed behavior. Runtime, network, CSP, and security claims still need visible diagnostics.

How this review workflow works

ClawSweeper keeps one durable marker-backed review comment per issue or PR.
Re-runs edit this comment so the latest verdict, findings, and automation markers stay together instead of adding duplicate bot comments.
A fresh review can be triggered by eligible @clawsweeper re-review comments, exact-item GitHub events, scheduled/background review runs, or manual workflow dispatch.
PR/issue authors and users with repository write access can comment @clawsweeper re-review or @clawsweeper re-run on an open PR or issue to request a fresh review only.
Maintainers can also comment @clawsweeper review to request a fresh review only.
Fresh-review commands do not start repair, autofix, rebase, CI repair, or automerge.
Maintainer-only repair and merge flows require explicit commands such as @clawsweeper autofix, @clawsweeper automerge, @clawsweeper fix ci, or @clawsweeper address review.
Maintainers can comment @clawsweeper explain to ask for more context, or @clawsweeper stop to stop active automation.

Review history (4 earlier review cycles)

reviewed 2026-07-01T17:25:19.463Z sha 8dc0cb2 :: needs real behavior proof before merge. :: [P2] Align StepFun input metadata with routed capabilities
reviewed 2026-07-09T05:34:43.912Z sha 8616240 :: needs maintainer review before merge. :: none
reviewed 2026-07-09T05:44:11.232Z sha 8616240 :: needs changes before merge. :: [P2] Run the docs formatter on the StepFun tables
reviewed 2026-07-09T06:04:37.591Z sha 6d7386f :: needs maintainer review before merge. :: none

clawsweeper · 2026-06-27T10:41:03Z

@lit26 thanks for the PR. ClawSweeper is still waiting on real behavior proof before this can move forward.

Useful proof can be a screenshot, short video, terminal output, copied live output, linked artifact, or redacted logs that show the changed behavior after the fix. Please redact private tokens, phone numbers, private endpoints, customer data, and anything else sensitive.

Once proof is added to the PR body or a comment, ClawSweeper or a maintainer can re-check it.

Output was artificially capped at 65536; StepFun docs set max_tokens default to INF, bounded only by the 256K context window. Raise step-3.7-flash maxTokens to 262144 on both stepfun and stepfun-plan, matching contextWindow. Per models.dev anomalyco/models.dev#1903. step-3.5-flash entries unchanged.

Step Plan endpoint bills per subscription plan, not per token, matching the sibling stepfun-plan/step-3.5-flash and step-3.5-flash-2603 entries which already carry zero cost. Standard stepfun/step-3.7-flash keeps its per-token rate. Per models.dev anomalyco/models.dev#1903, whose step-plan providers omit cost entirely.

steipete · 2026-07-09T06:06:04Z

Maintainer fixups complete and land-ready:

preserved the existing Step 3.5 defaults on Standard and Step Plan;
limited advertised OpenClaw input to text/image;
added low/medium/high reasoning propagation, including the agent-turn off -> low fallback;
matched StepFun's Chat Completions request contract (system, max_tokens, no OpenAI-only store/developer/strict/stream-options fields);
added emitted-payload regression coverage and provider docs; release-note context stays in the PR body because CHANGELOG.md is release-owned.

Exact-head proof at 0ad970c66111a7974954c85e6f54afcf7248dd6e:

sanitized public-AWS Crabbox run_564f467801fc (no secrets, instance role, tailnet, or LAN access);
targeted oxfmt --check: pass;
pnpm test extensions/stepfun/index.test.ts src/agents/models-config.providers.stepfun.test.ts: 6 tests pass across 2 shards;
fresh Codex autoreview: clean, no accepted/actionable findings.

Proof gap: no approved StepFun credential was available for direct inference. The emitted request payload is regression-tested against StepFun's official API contract linked in the PR body.

Thanks @lit26 for adding the model coverage.

steipete · 2026-07-09T06:14:51Z

Merged via squash.

Prepared head SHA: 0ad970c66111a7974954c85e6f54afcf7248dd6e
Landed commit: 583ca34bf88d86846c6ae26fd5e3c8cce7148a5c

* feat(stepfun): add step-3.7-flash model and make it default * fix(stepfun): set step-3.7-flash cacheRead cost to 0.04 * fix(stepfun): raise step-3.7-flash maxTokens to context limit Output was artificially capped at 65536; StepFun docs set max_tokens default to INF, bounded only by the 256K context window. Raise step-3.7-flash maxTokens to 262144 on both stepfun and stepfun-plan, matching contextWindow. Per models.dev anomalyco/models.dev#1903. step-3.5-flash entries unchanged. * fix(stepfun): zero step-plan step-3.7-flash cost for plan billing Step Plan endpoint bills per subscription plan, not per token, matching the sibling stepfun-plan/step-3.5-flash and step-3.5-flash-2603 entries which already carry zero cost. Standard stepfun/step-3.7-flash keeps its per-token rate. Per models.dev anomalyco/models.dev#1903, whose step-plan providers omit cost entirely. * docs(changelog): note StepFun 3.7 support * style(stepfun): format provider table * fix(stepfun): forward reasoning effort * fix(stepfun): scope reasoning compat to 3.7 * test(stepfun): keep 3.5 reasoning compat unchanged * fix(stepfun): carry off fallback through agent turns * fix(stepfun): use documented token limit field * fix(stepfun): match chat completions contract * chore(stepfun): keep changelog release-owned --------- Co-authored-by: Tianning Li <[email protected]> Co-authored-by: Peter Steinberger <[email protected]>

clawsweeper Bot added rating: 🦐 gold shrimp Decent PR readiness signal, but merge confidence is limited. and removed rating: 🧂 unranked krab Not merge-ready due to missing proof or serious correctness/safety concerns. labels May 29, 2026

clawsweeper Bot added rating: 🦪 silver shellfish Thin PR readiness signal; proof, validation, or implementation needs work. and removed rating: 🦐 gold shrimp Decent PR readiness signal, but merge confidence is limited. labels Jun 15, 2026

Tianning Li added 2 commits June 29, 2026 17:21

feat(stepfun): add step-3.7-flash model and make it default

aba94ce

fix(stepfun): set step-3.7-flash cacheRead cost to 0.04

ba5d6cc

lit26 force-pushed the feat/stepfun-step-3.7-flash branch from 3179016 to ba5d6cc Compare June 29, 2026 09:42

Tianning Li added 2 commits June 29, 2026 20:04

lit26 marked this pull request as ready for review June 30, 2026 16:39

style(stepfun): format provider table

ac8125e

steipete added 3 commits July 8, 2026 22:40

fix(stepfun): forward reasoning effort

6702153

fix(stepfun): scope reasoning compat to 3.7

07ceeb7

test(stepfun): keep 3.5 reasoning compat unchanged

9338be6

steipete added 3 commits July 8, 2026 22:49

fix(stepfun): carry off fallback through agent turns

e1ff08f

fix(stepfun): use documented token limit field

975b68c

fix(stepfun): match chat completions contract

6d7386f

chore(stepfun): keep changelog release-owned

0ad970c

steipete merged commit 583ca34 into openclaw:main Jul 9, 2026
105 checks passed

This was referenced Jul 9, 2026

fix(infra): drop replacement characters when response snippets truncate UTF-8 #102532

Closed

feat(stepfun): add video understanding (describeVideo) support #102181

Open

lit26 mentioned this pull request Jul 10, 2026

feat(stepfun): default to step-3.7-flash #103623

Open

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Uh oh!

feat(stepfun): add step-3.7-flash model#88082

feat(stepfun): add step-3.7-flash model#88082
steipete merged 15 commits into
openclaw:mainfrom
lit26:feat/stepfun-step-3.7-flash

lit26 commented May 29, 2026 •

edited by steipete

Loading

Uh oh!

clawsweeper Bot commented May 29, 2026 •

edited

Loading

Uh oh!

clawsweeper Bot commented Jun 27, 2026

Uh oh!

steipete commented Jul 9, 2026 •

edited

Loading

Uh oh!

Uh oh!

steipete commented Jul 9, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

Uh oh!

Uh oh!

Conversation

lit26 commented May 29, 2026 • edited by steipete Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

What Problem This Solves

Why This Change Was Made

User Impact

Evidence

Uh oh!

clawsweeper Bot commented May 29, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

clawsweeper Bot commented Jun 27, 2026

Uh oh!

steipete commented Jul 9, 2026 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

steipete commented Jul 9, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

2 participants

lit26 commented May 29, 2026 •

edited by steipete

Loading

clawsweeper Bot commented May 29, 2026 •

edited

Loading

steipete commented Jul 9, 2026 •

edited

Loading