[oyi77]

Summary

This PR addresses the 4th highest priority massive file from Issue #3594: open-sse/services/combo.ts at 4,600+ lines.

Changes

Splits the monolithic core combination service into 12 domain-focused files inside open-sse/services/combo/:

File	Contents
types.ts	All shared interfaces, runtime step types, target types
utils.ts	Validation helpers, normalizers, basic parsing
constants.ts	Shared weights, quotas, reset constants
state.ts	In-memory atomic counters, connection caches, soft penalties
dag.ts	DAG validation, nested target resolution, target flattening
sorting.ts	Target sorting (Cost-optimized, Least-used, Context-size, Weighted)
quota.ts	Reset-aware fetching, reset window alignment, and SLA scoring
context.ts	Context affinity, prompt extraction, and intent classification
shadow.ts	Shadow routing clones, execution, and draining
auto.ts	buildAutoCandidates scoring logic, request-tag routing
roundRobin.ts	The handleRoundRobinCombo execution path
chat.ts	The main handleComboChat execution orchestration loop
index.ts	Exposes the 15 exported functions back to the rest of the app

Results

• combo.ts: 4,657 lines → 1 line (re-exports combo/index.ts)
• All imports between files are resolved correctly.
• Full backward compatibility: all 15 previously exported functions from combo.ts are re-exported.

Verification

• npm run typecheck:core — passes cleanly.
• npx eslint open-sse/services/combo/*.ts — 0 errors or warnings.
• Strictly structural refactoring with identical runtime behavior.

Resolves part of #3594.

[gemini-code-assist]

Code Review

This pull request introduces several bug fixes, refactorings, and feature additions across the repository. Key changes include soft-deprecating @omniroute/opencode-provider in favor of @omniroute/opencode-plugin, modularizing the handleChatCore handler and the combo routing engine, and resolving several undocumented or missing API endpoints (such as gamification, guardrails, and agent-bridge). Additionally, it hardens the Antigravity/agy provider limits, translates chat-pipeline responses back to the legacy text-completion shape for /v1/completions, and migrates public OAuth client IDs to secure public credential resolvers. Feedback on the changes highlights a potential issue in open-sse/services/combo/constants.ts where an invalid QUOTA_SOFT_DEPRIORITIZE_FACTOR environment variable could parse to NaN and break sorting; a fallback check using Number.isFinite is recommended.

Important

The consumer version of Gemini Code Assist on GitHub is being sunset. Starting June 18, 2026, new organization installations will be blocked, and all code review activity will officially cease on July 17, 2026.
For more details on the timeline and next steps, please review the Help Documentation.

[gemini-code-assist]

If process.env.QUOTA_SOFT_DEPRIORITIZE_FACTOR is set to an invalid numeric string, Number(...) will return NaN. This will result in NaN scores during auto-combo scoring and break the sorting logic. Use Number.isFinite to safely fall back to 0.7 if the parsed value is invalid.

const rawFactor = Number(process.env.QUOTA_SOFT_DEPRIORITIZE_FACTOR ?? "0.7");
export const QUOTA_SOFT_DEPRIORITIZE_FACTOR = Number.isFinite(rawFactor) ? rawFactor : 0.7;

[diegosouzapw]

Thanks @oyi77 for the modularization work on open-sse/services/combo.ts (Issue #3594). I can't merge this as it currently stands — and the reason is structural, not about the quality of the split:

1. It's a stacked/cumulative chain. This branch carries ~49–55 commits all starting from chore(release): open v3.8.23 development cycle, and the per-PR additions grow monotonically (refactor: modularize providerRegistry.ts (Issue #3594) #3617 ≈ +22k → refactor: modularize 3.1K-line providers.ts constants (Issue #3594) #3646 ≈ +52k). Each refactor branch is built on top of the previous one, so merging any single PR drags in all the earlier refactors too. That makes each PR impossible to review in isolation and guarantees conflicts between them.

2. The base is stale (v3.8.23 era). Since these were cut, release/v3.8.24 has merged dozens of PRs — including changes to the very files several of these PRs modularize (e.g. providerRegistry.ts grew via fix(providers): correct Ollama Cloud kimi-k2.7-code capabilities (#3761) #3768, validation.ts via fix(providers): repair qwen-web validation + stop mislabeling benign redirects as SSRF (#3288, #3758) #3767, plus combo/stream/usage edits). GitHub now reports this PR as CONFLICTING, and the conflicts are substantial.

What would let this land: re-cut it as a single, independent PR branched from the current release/v3.8.24 tip, modularizing only open-sse/services/combo.ts (not stacked on the other refactors), small enough to review on its own. One file per PR, rebased on current release. I'll gladly review and merge them in that form — the god-file modularization goal is worth it; it's the 14-deep stack on a 2-week-old base that can't go in. Leaving this open so nothing is lost.

[oyi77]

Closed in favor of clean re-cut — see #3793 (providerRegistry), #3798 (combo), #3797 (validation), #3796 (imageGeneration), #3795 (usage), #3794 (providers). Only modularization commits were cherry-picked onto current release/v3.8.24 — no stacked dependencies, no unrelated changes.