Files
OmniRoute/docs/compression/2026-06-21-compression-phase3-request-header-plan.md
Diego Rodrigues de Sa e Souza 19d91d82e2 Release v3.8.34 (#4614)
* chore(release): open v3.8.34 development cycle

* chore(quality): release-green pre-flight validator + nightly signal (C+D) (#4622)

C — scripts/quality/validate-release-green.mjs (npm run check:release-green):
reproduces the release-equivalent validation (typecheck, eslint, db-rules,
public-creds, full unit, vitest, ratchets, optional --with-build package-artifact)
against the current working tree and classifies each red as HARD (real defect,
exit 1) vs DRIFT (ratchet — reported, never affects exit / never blocks). Pure
helpers exported + orchestration behind a direct-run guard; unit-tested.

D — .github/workflows/nightly-release-green.yml: runs C on the active release
branch nightly (and on workflow_dispatch) and opens/updates a single tracking
issue on HARD failures. Never a required check, never touches a contributor PR.

Closes the gap where the full gate (ci.yml) only ran on the release PR, so reds
accrued silently on release/** and surfaced in 40-min layers at release time.
Non-blocking by construction; drift is the maintainer's to rebaseline at release.

Co-authored-by: Diego Rodrigues de Sa e Souza <diego.souza@cdwasolutions.com.br>

* fix(providers): show revealed connection API keys (#4583)

Integrated into release/v3.8.34

* fix(resilience): respect upstream retry hint toggle (#4585)

Integrated into release/v3.8.34

* feat(settings): expose stream recovery feature flags (#4586)

Integrated into release/v3.8.34

* fix(logs): make active request stale sweep configurable (#4599)

Integrated into release/v3.8.34

* fix(plugin): auto-prefix providerId with 'opencode-' for OC 1.17.8+ native gate (#4527)

Integrated into release/v3.8.34 (supersedes #4445)

* fix(models): treat unknown output caps as unset (#4584)

Integrated into release/v3.8.34

* fix(executors): strip temperature for GitHub Copilot gpt-5.4 family (#4564)

Integrated into release/v3.8.34 (rebuilt onto tip)

* fix(oauth): update Qwen OAuth URLs from chat.qwen.ai to qwen.ai (#4561)

Integrated into release/v3.8.34 (rebuilt onto tip)

* fix(api/settings): prevent cached /api/settings responses (port from 9router#951) (#4566)

Integrated into release/v3.8.34 (rebuilt onto tip)

* feat(audio): MiniMax T2A v2 TTS dispatch in audioSpeech (port #1043) (#4553)

Integrated into release/v3.8.34 (rebuilt onto tip)

* fix(dashboard): surface manual config CTA when Open Claw CLI auto-detect fails (#4562)

Integrated into release/v3.8.34 (rebuilt onto tip)

* feat(providers): optional model ID for custom API-key validation (#4555)

Integrated into release/v3.8.34 (rebuilt onto tip)

* fix(cli): align data dir and env loading with runtime (#4607)

Integrated into release/v3.8.34 (rebuilt onto tip)

* fix(quota): expose Bailian quota windows (#4610)

Integrated into release/v3.8.34 (rebuilt onto tip)

* fix: retain provider cooldowns for configured max window (#4588)

Integrated into release/v3.8.34 (rebuilt — bundled commits stripped)

* fix: reject invalid provider cooldown bounds (#4589)

Integrated into release/v3.8.34 (rebuilt — bundled commits stripped)

* fix: preserve production combo metrics on shadow eviction (#4590)

Integrated into release/v3.8.34 (rebuilt — bundled commits stripped)

* fix(stream): estimate input tokens when upstream reports prompt_tokens=0 (#4615)

Integrated into release/v3.8.34 (rebuilt onto tip)

* fix(catalog): shorten no-thinking gateway prefix to no-think/ (#4525)

Integrated into release/v3.8.34 (rebuilt — kept only the prefix rename, dropped stale-base reverts)

* fix(relay): apply IP rate limit to bifrost sidecar (#4593)

Integrated into release/v3.8.34 (rebuilt onto tip; merge before #4612)

* fix(bifrost): finalize SSE relay usage after stream (#4612)

Integrated into release/v3.8.34 (rebuilt + reconciled with #4593)

* feat(compression): per-request `x-omniroute-compression` header (Phase 3) (#4645)

* docs(compression): Phase 3 per-request header design spec

Approved brainstorming output for the x-omniroute-compression header:
header-first precedence, name-first combo matching (Decision A), explicit
value bypasses auto-trigger (Decision B), DerivedPlan.source, and the
X-OmniRoute-Compression response header.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* docs(compression): Phase 3 per-request header implementation plan

4-task TDD plan (resolver header-first + source, parser, chatCore wiring +
response header, docs/file-size) with full code and exact commands.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* feat(compression): header-first resolver + plan source (Phase 3 core)

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* feat(compression): resolveCompressionHeader parser (Phase 3)

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* feat(compression): wire x-omniroute-compression header + response header (Phase 3)

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* refactor(compression): extract plan-resolution leaf (planResolution.ts) under size cap (Phase 3)

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* docs(compression): document x-omniroute-compression header (Phase 3)

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

* fix(compression): harden named-combo map + trim engine: header id (Phase 3 review)

Addresses gemini-code-assist review on #4645:
- Extract buildNamedComboLookup (pure) so a blank/whitespace/null combo name
  contributes only its id key (no '' key, no throw that disables all combos).
- Trim the engine:<id> header value so 'engine: rtk' resolves.

Co-Authored-By: Claude Opus 4.8 (1M context) <noreply@anthropic.com>

---------

Co-authored-by: Diego Rodrigues de Sa e Souza <diego.souza@cdwasolutions.com.br>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-authored-by: Diego Rodrigues de Sa e Souza <souzamiriamrodrigues790@gmail.com>

* fix: exclude exhausted connections from auto scoring (#4592)

Integrated into release/v3.8.34 (rebuilt + opt-in gate fix)

* fix(dashboard): memoize compatible provider groups (#4613)

Integrated into release/v3.8.34 (rebuilt + test added)

* fix(dashboard): isolate quota widget refresh clock (#4611)

Integrated into release/v3.8.34 (rebuilt + jsdom test)

* fix(dashboard): gate topology side effects behind widget visibility (#4606)

Integrated into release/v3.8.34 (rebuilt + jsdom test)

* fix(dashboard): keep play_arrow spinning on provider Test All buttons (#4563)

Integrated into release/v3.8.34 (rebuilt onto tip; UI-cosmetic per owner)

* fix(db): schedule retention cleanup + fix cleanup table/column names (extracted from #4428) (#4691)

Integrated into release/v3.8.34 (cleanup core extracted from #4428, credit @oyi77)

* fix(telemetry): back off live-WS event forwarding when the sidecar is unreachable (#4604) (#4687)

Co-authored-by: Diego Rodrigues de Sa e Souza <souzamiriamrodrigues790@gmail.com>

* fix(api): serve GET /v1/models/{model} as JSON, not the HTML dashboard (#4674) (#4677)

Co-authored-by: Diego Rodrigues de Sa e Souza <souzamiriamrodrigues790@gmail.com>

* feat(opencode): add go deepseek reasoning variants (#4647)

Integrated into release/v3.8.34

* fix(executors): robust deepseek-web tool-call parsing and agentic context retention (#4644)

Integrated into release/v3.8.34

* fix(cli): authenticate `omniroute logs` and honor active context (#4638)

Integrated into release/v3.8.34 (authored by Rahul Sharma, AI co-author trailer stripped per project policy)

* fix(proxy): apply pipelining:0 + connections cap to the direct dispatcher (#4580) (#4684)

Co-authored-by: Diego Rodrigues de Sa e Souza <souzamiriamrodrigues790@gmail.com>

* fix(executors): Firecrawl web_fetch 500 with include_metadata=true (#4692)

Integrated into release/v3.8.34

* fix(routing): include all noAuth models in auto-combos + add reka-flash + best-free template (#4621)

Integrated into release/v3.8.34 (dead getFirstRegistryModelId dropped, rebuilt onto tip)

* fix(dashboard): gate home topology live-WS networking (#4596) (#4618)

Integrated into release/v3.8.34 (adapted onto #4606's extracted topology section: default-hidden flip + enabled gate on useLiveDashboard)

* fix(cli): align `omniroute` env loading with the runtime data dir (#4597) (#4619)

Integrated into release/v3.8.34 (data-dir.mjs refactor reconciled with #4607; loadEnvFile aligned to getDefaultDataDir)

* chore(quality): reconcile file-size baseline for #4644 (deepseek-web.ts 1117->1125) (#4695)

file-size reconcile for #4644

* Support quota scraping for OpenCode Go and Ollama Cloud (#4642)

Integrated into release/v3.8.34 (Ollama Cloud + OpenCode Go dashboard quota scraping; rebuilt onto tip, gates green: typecheck/public-creds/file-size/lint/docs-sync + 31 tests)

* feat(executors): land M365 Copilot pure framing + connection helpers (#4042) (#4696)

Land M365 pure modules ahead of draft #4400

* deps: bump production + development groups; migrate js-yaml to v5 ESM (#4697)

Incorporates Dependabot #4667 + #4668 + js-yaml v5 ESM migration into release/v3.8.34

* fix: noAuth provider validation + kimi executor routing (#4699)

Integrated into release/v3.8.34 (noAuth in NOAUTH_PROVIDERS dynamic check + remove misrouted kimi web alias; 9 tests)

* refactor(imageGeneration): extract 8 provider families to co-located files (#4609)

Integrated into release/v3.8.34 (extraction completed: added missing imports/exports per module, main imports handlers locally; 145 image-gen tests pass, typecheck/cycles/file-size green)

* chore(release): v3.8.34 — finalize changelog, rebaseline drift, fix release-green reds

- Finalize CHANGELOG [3.8.34] (43 bullets, full contributor attribution) + seed i18n mirrors
- Rebaseline inherited cycle drift surfaced by release-green pre-flight: eslint warnings
  3900->3907, cognitive-complexity 797->801 (release-finalize touches no prod code; all
  drift is from this cycle's contributor merges)
- fix(providers): keep reka-flash-3 as the Reka provider default. #4621 inserted reka-flash
  at the head of the model list, silently changing the default from reka-flash-3 (the
  free-tier model) to reka-flash; reorder so reka-flash-3 stays default, reka-flash retained.
- test: align provider-models-config / provider-models-route / web-cookie-providers-new with
  #4621 (reka-flash now in the Reka catalog) and #4699 (the `kimi` API-key provider correctly
  falls through to DefaultExecutor instead of KimiWebExecutor)
- chore(quality): allowlist the COMPRESSION_GUIDE doc name in check-fabricated-docs
  (false-positive env-var match; docs/compression/COMPRESSION_GUIDE.md exists)

* fix(release-green): resolve release-PR full-CI reds for v3.8.34

Surfaced only on the release PR (these gates don't run on PR->release fast-gates):

- fix(quota): complete HTML-comment sanitization in opencodeOllamaUsage SSR reset-time
  parsing — strip any <!--...--> generically instead of the two literal React hydration
  markers, so no partial "<!--" can survive (CodeQL js/incomplete-multi-character-
  sanitization, HIGH, introduced by #4642). Regression test added.
- test(codex): correct the Codex-fingerprint body key order assertion to match the
  canonical bodyFieldOrder (prompt_cache_key precedes include); #4584 flipped the two
  and integration tests don't run on fast-gates so it never executed until the release PR.
- chore(quality): rebaseline inherited cycle drift surfaced by full CI —
  zizmorFindings 152->155 (+3 unpinned-uses in nightly-release-green.yml from #4622,
  same @vN convention as ci.yml) and openapiCoverage.pct 38.4->37.8 (-0.6, contributor
  routes added faster than openapi docs). Release-finalize touches no prod routes.

* fix(release-green): complete CodeQL sanitization + rebaseline complexity drift

- fix(quota): handle unterminated HTML comments in opencodeOllamaUsage SSR reset-time
  parsing — the `(?:-->|$)` arm consumes a trailing "<!--" with no closing "-->", so no
  partial "<!--" can survive (CodeQL js/incomplete-multi-character-sanitization persisted
  with the plain <!--...--> form because an unclosed comment could still leave "<!--").
- chore(quality): rebaseline cyclomatic complexity 1915->1916 (+1) — inherited v3.8.34
  cycle drift (contributor feature branches); check:complexity does not run on PR->release
  fast-gates so it surfaced only on the release PR. Release-finalize adds 0 complexity
  (measured 1916 with/without the regex tweak). dead-code/cognitive/type-coverage/
  compression-budget/codeql ratchets all pass.

---------

Co-authored-by: Diego Rodrigues de Sa e Souza <diego.souza@cdwasolutions.com.br>
Co-authored-by: Randi <55005611+rdself@users.noreply.github.com>
Co-authored-by: Hernan Javier Ardila Sanchez <hjasgr@gmail.com>
Co-authored-by: KooshaPari <42529354+KooshaPari@users.noreply.github.com>
Co-authored-by: Abhishek Divekar <adivekar@utexas.edu>
Co-authored-by: Rahul sharma <sharmaR0810@gmail.com>
Co-authored-by: Claude Opus 4.8 (1M context) <noreply@anthropic.com>
Co-authored-by: Diego Rodrigues de Sa e Souza <souzamiriamrodrigues790@gmail.com>
Co-authored-by: Ronald Estacion <DevEstacion@users.noreply.github.com>
Co-authored-by: Igor <60442260+BugsBag@users.noreply.github.com>
Co-authored-by: Oonishi <275808243+ponkcore@users.noreply.github.com>
Co-authored-by: Paijo <14921983+oyi77@users.noreply.github.com>
Co-authored-by: Jan Leon <Jan.gaschler@gmail.com>
2026-06-23 03:08:29 -03:00

32 KiB

title, version, lastUpdated
title version lastUpdated
Compression Phase 3: Per-Request Header — Implementation Plan 3.8.33 2026-06-21

Compression Phase 3: Per-Request Header — Implementation Plan

For agentic workers: REQUIRED SUB-SKILL: Use superpowers:subagent-driven-development (recommended) or superpowers:executing-plans to implement this plan task-by-task. Steps use checkbox (- [ ]) syntax for tracking.

Goal: Add the x-omniroute-compression per-request header so a client can override the resolved compression plan for a single request, taking precedence over every operator-configured layer.

Architecture: A pure header interpreter (planFromHeader) is evaluated at the top of the compression resolver (resolveBasePlan in strategySelector.ts); chatCore parses the header off the wire, threads it through the existing selector signature (the same way Phase 2 threaded combos), captures the resolved { mode, source }, and emits an X-OmniRoute-Compression response header. The resolver stays pure (no DB import).

Tech Stack: TypeScript, Node test runner (node --import tsx/esm --test), better-sqlite3 (combos store, read via chatCore only), Next.js handler (open-sse/handlers/chatCore.ts).

Spec: docs/compression/2026-06-21-compression-phase3-request-header-design.md


Key prior-art / facts (read before starting)

  • The resolver entry point is resolveBasePlan in open-sse/services/compression/strategySelector.ts. It already threads a combos map (Phase 2). Public selectors: selectCompressionPlan / selectCompressionStrategy / getEffectiveMode, signature ending in ..., combos = {}.
  • DerivedPlan is defined in open-sse/services/compression/deriveDefaultPlan.ts as { mode: string; stackedPipeline: Array<{ engine: string; intensity?: string }> }.
  • open-sse/services/compression/resolveCompressionPlan.ts carries a dormant header branch (ctx.header + private headerToPlan). No caller ever passes header (confirmed: the only .header refs in compression code are those lines; no test passes it). This plan removes that dormant branch so header interpretation has a single home.
  • Parsing precedent: isNoMemoryRequested in open-sse/handlers/chatCore/headers.ts (uses getHeaderValueCaseInsensitive). Tests for that module: tests/unit/chatcore-headers.test.ts.
  • A combo's id is a uuidv4() (or the seeded slug default-caveman); name has no UNIQUE constraint (migration 042). Header <combo> matches name-first (Decision A).
  • chatCore reads request headers from clientRawRequest?.headers. The compression block builds namedCombos (~line 1373) and calls selectCompressionStrategy (~line 1532) then selectCompressionPlan (~line 1601). Response headers are built at two sites: the non-streaming JSON path (responseHeaders at ~line 4570, after attachOmniRouteMetaHeaders) and the streaming path (responseHeaders at ~line 4697).
  • Decision A: <combo> matched name-first (lowercased), then exact id. Decision B: any valid header value bypasses auto-trigger.
  • Boundary (implement as written): the master switch is the hard gate — when config.enabled is false, the request is uncompressed regardless of the header (the header redirects among configured plans only when compression is enabled). A test asserts this.

File Structure

File Responsibility Change
open-sse/services/compression/deriveDefaultPlan.ts DerivedPlan type + new CompressionSource union + optional source. Modify
open-sse/services/compression/strategySelector.ts planFromHeader interpreter, header-first precedence in resolveBasePlan, source tagging, formatCompressionMeta, thread header through public selectors. Modify
open-sse/services/compression/resolveCompressionPlan.ts Remove the dormant header branch + headerToPlan + ResolveCtx.header. Modify
open-sse/handlers/chatCore/headers.ts resolveCompressionHeader parser. Modify
src/shared/constants/headers.ts compression response-header name. Modify
open-sse/handlers/chatCore.ts Parse header, add name keys to namedCombos, thread header, capture {mode,source}, emit response header at both sites. Modify
tests/unit/compression/compression-header-dispatch.test.ts Resolver + planFromHeader + source + precedence + formatCompressionMeta. Create
tests/unit/chatcore-headers.test.ts resolveCompressionHeader parsing. Modify
docs/reference/API_REFERENCE.md, docs/compression/COMPRESSION_GUIDE.md Document the header. Modify
config/quality/file-size-baseline.json Rebaseline chatCore if it crosses its pin. Modify (if needed)

Task 1: Header interpreter, precedence, and source in the resolver

Files:

  • Modify: open-sse/services/compression/deriveDefaultPlan.ts

  • Modify: open-sse/services/compression/strategySelector.ts

  • Modify: open-sse/services/compression/resolveCompressionPlan.ts

  • Test: tests/unit/compression/compression-header-dispatch.test.ts (create)

  • Step 1: Write the failing test

Create tests/unit/compression/compression-header-dispatch.test.ts:

import { describe, it } from "node:test";
import assert from "node:assert/strict";
import {
  selectCompressionStrategy,
  selectCompressionPlan,
  planFromHeader,
  formatCompressionMeta,
} from "../../../open-sse/services/compression/strategySelector.ts";
import {
  DEFAULT_COMPRESSION_CONFIG,
  type CompressionConfig,
} from "../../../open-sse/services/compression/types.ts";

const combos = {
  c1: [{ engine: "rtk", intensity: "standard" }, { engine: "caveman", intensity: "full" }],
  "fast combo": [{ engine: "lite" }],
};

function cfg(overrides: Partial<CompressionConfig> = {}): CompressionConfig {
  return { ...DEFAULT_COMPRESSION_CONFIG, enabled: true, ...overrides };
}

// header is the 7th positional arg of selectCompressionPlan/selectCompressionStrategy.
function planWithHeader(config: CompressionConfig, header: string | null) {
  return selectCompressionPlan(config, null, 0, undefined, undefined, combos, header);
}

describe("planFromHeader (Phase 3)", () => {
  it("off => mode off, source request-header", () => {
    const p = planFromHeader(cfg(), "off", combos);
    assert.deepEqual(p, { mode: "off", stackedPipeline: [], source: "request-header" });
  });

  it("default => the panel-derived default, ignoring the active profile", () => {
    const config = cfg({
      activeComboId: "c1",
      enginesExplicit: true,
      engines: { rtk: { enabled: true } },
    });
    const p = planFromHeader(config, "default", combos);
    assert.equal(p?.mode, "rtk"); // engines map default, NOT the c1 active profile
    assert.equal(p?.source, "request-header");
  });

  it("engine:<id> => that single engine when enabled (case-insensitive)", () => {
    const config = cfg({ enginesExplicit: true, engines: { rtk: { enabled: true } } });
    assert.equal(planFromHeader(config, "engine:RTK", combos)?.mode, "rtk");
  });

  it("engine:<id> => null (fall-through) when the engine is disabled", () => {
    const config = cfg({ engines: { rtk: { enabled: false } } });
    assert.equal(planFromHeader(config, "engine:rtk", combos), null);
  });

  it("<combo> matches by name (case-insensitive) and by id", () => {
    assert.deepEqual(planFromHeader(cfg(), "FAST COMBO", combos)?.stackedPipeline, combos["fast combo"]);
    assert.deepEqual(planFromHeader(cfg(), "c1", combos)?.stackedPipeline, combos.c1);
  });

  it("unknown value => null (fall-through)", () => {
    assert.equal(planFromHeader(cfg(), "nonsense", combos), null);
  });
});

describe("header precedence in resolveBasePlan (Phase 3)", () => {
  it("a valid header beats the active profile", () => {
    const config = cfg({ activeComboId: "c1" });
    assert.equal(planWithHeader(config, "off").mode, "off");
    assert.equal(planWithHeader(config, "off").source, "request-header");
  });

  it("a valid header beats a routing-combo override", () => {
    const config = cfg({ comboOverrides: { "route-x": "stacked" } });
    const plan = selectCompressionPlan(config, "route-x", 0, undefined, undefined, combos, "off");
    assert.equal(plan.mode, "off");
    assert.equal(plan.source, "request-header");
  });

  it("a valid header bypasses auto-trigger (Decision B)", () => {
    const config = cfg({
      autoTriggerTokens: 1000,
      autoTriggerMode: "aggressive",
      enginesExplicit: true,
      engines: { rtk: { enabled: true } },
    });
    // Large prompt would auto-escalate to aggressive; the header pins the panel default.
    assert.equal(planWithHeader(config, "default").mode, "rtk");
  });

  it("an unknown header falls through to the normal resolution", () => {
    const config = cfg({ activeComboId: "c1" });
    const plan = planWithHeader(config, "bogus");
    assert.equal(plan.mode, "stacked");
    assert.equal(plan.source, "active-profile");
  });

  it("master-off beats the header (hard kill switch)", () => {
    const config = cfg({ enabled: false, engines: { rtk: { enabled: true } } });
    const plan = planWithHeader(config, "engine:rtk");
    assert.equal(plan.mode, "off");
    assert.equal(plan.source, "off");
  });
});

describe("source on non-header paths", () => {
  it("routing-override / active-profile / auto-trigger / default / off", () => {
    assert.equal(
      selectCompressionPlan(cfg({ comboOverrides: { r: "lite" } }), "r", 0, undefined, undefined, combos, null).source,
      "routing-override"
    );
    assert.equal(planWithHeader(cfg({ activeComboId: "c1" }), null).source, "active-profile");
    assert.equal(
      planWithHeader(cfg({ autoTriggerTokens: 10, autoTriggerMode: "lite" }), null).source,
      "auto-trigger"
    );
    assert.equal(
      planWithHeader(cfg({ enginesExplicit: true, engines: { rtk: { enabled: true } } }), null).source,
      "default"
    );
    assert.equal(planWithHeader(cfg({ enabled: false }), null).source, "off");
  });
});

describe("formatCompressionMeta", () => {
  it("renders '<mode>; source=<source>'", () => {
    assert.equal(
      formatCompressionMeta({ mode: "aggressive", stackedPipeline: [], source: "request-header" }),
      "aggressive; source=request-header"
    );
    assert.equal(formatCompressionMeta({ mode: "off", stackedPipeline: [] }), "off; source=off");
  });
});
  • Step 2: Run the test to verify it fails

Run: node --import tsx --import ./open-sse/utils/setupPolyfill.ts --import ./tests/_setup/isolateDataDir.ts --test --test-force-exit tests/unit/compression/compression-header-dispatch.test.ts Expected: FAIL — planFromHeader/formatCompressionMeta are not exported yet (import error / not a function).

  • Step 3: Add CompressionSource + source to DerivedPlan

In open-sse/services/compression/deriveDefaultPlan.ts, replace the DerivedPlan interface block:

export type CompressionSource =
  | "request-header"
  | "routing-override"
  | "active-profile"
  | "auto-trigger"
  | "default"
  | "off";

export interface DerivedPlan {
  mode: string;
  stackedPipeline: Array<{ engine: string; intensity?: string }>;
  /** Which precedence layer decided this plan (Phase 3 observability). Optional so
   *  Phase 1/2 callers and snapshots are unaffected. */
  source?: CompressionSource;
}

(Leave the deriveDefaultPlan function body unchanged — it does not set source; the resolver tags it.)

  • Step 4: Remove the dormant header branch from resolveCompressionPlan.ts

In open-sse/services/compression/resolveCompressionPlan.ts: delete the header field from ResolveCtx, delete the // 1. header block, and delete the private headerToPlan function. The file becomes:

import { deriveDefaultPlan, type DerivedPlan } from "./deriveDefaultPlan.ts";

export interface ResolveCtx {
  comboId?: string | null;
  combos?: Record<string, Array<{ engine: string; intensity?: string }>>; // named combo pipelines by id
}

export function resolveCompressionPlan(config: any, ctx: ResolveCtx): DerivedPlan {
  if (config?.enabled === false) return { mode: "off", stackedPipeline: [] };

  // routing-combo override
  const ov = ctx.comboId ? config?.comboOverrides?.[ctx.comboId] : undefined;
  if (ov) return modeToPlan(ov, config);

  // active named combo
  if (config?.activeComboId && ctx.combos?.[config.activeComboId]) {
    return { mode: "stacked", stackedPipeline: ctx.combos[config.activeComboId] };
  }

  // derived default
  return deriveDefaultPlan(config?.engines ?? {}, config?.enabled !== false);
}

function modeToPlan(mode: string, config: any): DerivedPlan {
  return mode === "stacked"
    ? { mode: "stacked", stackedPipeline: config?.stackedPipeline ?? [] }
    : { mode, stackedPipeline: [] };
}
  • Step 5: Implement planFromHeader, withSource, header-first precedence, and formatCompressionMeta in strategySelector.ts

In open-sse/services/compression/strategySelector.ts:

(a) Update the DerivedPlan import to also bring CompressionSource:

import { deriveDefaultPlan, type DerivedPlan, type CompressionSource } from "./deriveDefaultPlan.ts";

(b) Add these helpers (place them just above resolveBasePlan):

/** Tags a plan with the precedence layer that produced it (Phase 3 observability). */
function withSource(plan: DerivedPlan, source: CompressionSource): DerivedPlan {
  return { ...plan, source };
}

/**
 * Interprets the `x-omniroute-compression` request header into a plan, or null when the
 * value is unrecognized (caller falls through to normal resolution). Pure.
 *   off            -> no compression
 *   default        -> the panel-derived Default (ignores active profile / routing / auto-trigger)
 *   engine:<id>    -> that single engine, when enabled in config.engines
 *   <combo>        -> a named combo, matched name-first (lowercased) then exact id (Decision A)
 */
export function planFromHeader(
  config: CompressionConfig,
  header: string,
  combos: NamedCombos
): DerivedPlan | null {
  const h = header.trim();
  if (!h) return null;
  const lower = h.toLowerCase();

  if (lower === "off") return withSource({ mode: "off", stackedPipeline: [] }, "request-header");

  if (lower === "default") {
    // Empty combos + null comboId yields the pure panel default (no active-combo leak).
    return withSource(deriveDefaultPlanFromConfig(config, null, {}), "request-header");
  }

  if (lower.startsWith("engine:")) {
    const id = lower.slice("engine:".length);
    const engine = config.engines?.[id];
    return engine?.enabled
      ? withSource(deriveDefaultPlan({ [id]: engine }, true), "request-header")
      : null;
  }

  const combo = combos[lower] ?? combos[h];
  return combo ? withSource({ mode: "stacked", stackedPipeline: combo }, "request-header") : null;
}

/** Renders the X-OmniRoute-Compression response header value. */
export function formatCompressionMeta(plan: DerivedPlan): string {
  return `${plan.mode}; source=${plan.source ?? "off"}`;
}

(c) Add a header parameter to resolveBasePlan and tag every return with withSource:

function resolveBasePlan(
  config: CompressionConfig,
  comboId: string | null,
  estimatedTokens: number,
  combos: NamedCombos = {},
  header: string | null = null
): DerivedPlan {
  if (!config.enabled) return withSource({ mode: "off", stackedPipeline: [] }, "off");

  // Phase 3: an explicit, recognized header wins over every operator layer (Decision B).
  // The master switch above is the hard kill: a header cannot turn compression on.
  if (header) {
    const fromHeader = planFromHeader(config, header, combos);
    if (fromHeader) return fromHeader; // already tagged "request-header"
  }

  const comboMode = checkComboOverride(config, comboId);
  if (comboMode) {
    return withSource(resolveCompressionPlan(config, { comboId, combos }), "routing-override");
  }

  if (config.activeComboId && combos[config.activeComboId]) {
    return withSource(
      { mode: "stacked", stackedPipeline: combos[config.activeComboId] },
      "active-profile"
    );
  }

  if (shouldAutoTrigger(config, estimatedTokens)) {
    const mode = config.autoTriggerMode ?? "lite";
    return withSource(
      mode === "stacked"
        ? { mode, stackedPipeline: config.stackedPipeline ?? [] }
        : { mode, stackedPipeline: [] },
      "auto-trigger"
    );
  }

  const plan = deriveDefaultPlanFromConfig(config, comboId, combos);
  return withSource(plan, plan.mode === "off" ? "off" : "default");
}

(d) Thread header through the public selectors (append as the last positional arg, default null), preserving source through the caching-aware adjustment:

export function getEffectiveMode(
  config: CompressionConfig,
  comboId: string | null,
  estimatedTokens: number,
  combos: NamedCombos = {},
  header: string | null = null
): CompressionMode {
  return resolveBasePlan(config, comboId, estimatedTokens, combos, header).mode as CompressionMode;
}

export function selectCompressionPlan(
  config: CompressionConfig,
  comboId: string | null,
  estimatedTokens: number,
  body?: Record<string, unknown>,
  context?: CachingDetectionContext,
  combos: NamedCombos = {},
  header: string | null = null
): DerivedPlan {
  const plan = resolveBasePlan(config, comboId, estimatedTokens, combos, header);
  if (body) {
    const ctx = detectCachingContext(body, context);
    const cacheAware = getCacheAwareStrategy(plan.mode as CompressionMode, ctx);
    return { ...plan, mode: cacheAware.strategy as CompressionMode }; // ...plan preserves source
  }
  return plan;
}

export function selectCompressionStrategy(
  config: CompressionConfig,
  comboId: string | null,
  estimatedTokens: number,
  body?: Record<string, unknown>,
  context?: CachingDetectionContext,
  combos: NamedCombos = {},
  header: string | null = null
): CompressionMode {
  return selectCompressionPlan(config, comboId, estimatedTokens, body, context, combos, header)
    .mode as CompressionMode;
}
  • Step 6: Run the test to verify it passes

Run: node --import tsx --import ./open-sse/utils/setupPolyfill.ts --import ./tests/_setup/isolateDataDir.ts --test --test-force-exit tests/unit/compression/compression-header-dispatch.test.ts Expected: PASS (all cases).

  • Step 7: Run the full compression suite to confirm no regression

Run: node --import tsx --import ./open-sse/utils/setupPolyfill.ts --import ./tests/_setup/isolateDataDir.ts --test --test-force-exit "tests/unit/compression/**/*.test.ts" Expected: PASS — the existing 774+ compression tests stay green (resolveCompressionPlan / strategySelector / active-combo dispatch unchanged in behavior for header-less callers).

  • Step 8: Commit
git add open-sse/services/compression/deriveDefaultPlan.ts \
        open-sse/services/compression/strategySelector.ts \
        open-sse/services/compression/resolveCompressionPlan.ts \
        tests/unit/compression/compression-header-dispatch.test.ts
git commit -m "feat(compression): header-first resolver + plan source (Phase 3 core)"

Task 2: resolveCompressionHeader parser

Files:

  • Modify: open-sse/handlers/chatCore/headers.ts

  • Test: tests/unit/chatcore-headers.test.ts

  • Step 1: Write the failing test

Append to tests/unit/chatcore-headers.test.ts (keep existing imports; add resolveCompressionHeader to the import from ../../open-sse/handlers/chatCore/headers.ts):

import { resolveCompressionHeader } from "../../open-sse/handlers/chatCore/headers.ts";
import { describe, it } from "node:test";
import assert from "node:assert/strict";

describe("resolveCompressionHeader", () => {
  it("reads the raw value case-insensitively and trims it", () => {
    assert.equal(resolveCompressionHeader({ "x-omniroute-compression": "  engine:rtk " }), "engine:rtk");
    assert.equal(resolveCompressionHeader(new Headers({ "X-OmniRoute-Compression": "off" })), "off");
  });

  it("returns null when absent or blank", () => {
    assert.equal(resolveCompressionHeader({}), null);
    assert.equal(resolveCompressionHeader({ "x-omniroute-compression": "   " }), null);
    assert.equal(resolveCompressionHeader(null), null);
  });
});
  • Step 2: Run the test to verify it fails

Run: node --import tsx --import ./open-sse/utils/setupPolyfill.ts --import ./tests/_setup/isolateDataDir.ts --test --test-force-exit tests/unit/chatcore-headers.test.ts Expected: FAIL — resolveCompressionHeader is not exported.

  • Step 3: Implement the parser

Append to open-sse/handlers/chatCore/headers.ts:

/**
 * Per-request compression override via the `x-omniroute-compression` header. Mirrors the
 * `x-omniroute-no-memory` convention (#4290). Returns the raw trimmed value, or null when
 * absent/blank. The resolver (planFromHeader) owns interpretation and casing rules; this
 * helper only reads the wire.
 */
export function resolveCompressionHeader(
  headers: Record<string, unknown> | Headers | null | undefined
): string | null {
  const value = (getHeaderValueCaseInsensitive(headers, "x-omniroute-compression") || "").trim();
  return value || null;
}
  • Step 4: Run the test to verify it passes

Run: node --import tsx --import ./open-sse/utils/setupPolyfill.ts --import ./tests/_setup/isolateDataDir.ts --test --test-force-exit tests/unit/chatcore-headers.test.ts Expected: PASS.

  • Step 5: Commit
git add open-sse/handlers/chatCore/headers.ts tests/unit/chatcore-headers.test.ts
git commit -m "feat(compression): resolveCompressionHeader parser (Phase 3)"

Task 3: chatCore wiring + response header

Files:

  • Modify: src/shared/constants/headers.ts

  • Modify: open-sse/handlers/chatCore.ts

  • Step 1: Add the response-header name constant

In src/shared/constants/headers.ts, add the compression entry to OMNIROUTE_RESPONSE_HEADERS (alphabetical, after cacheHit):

  cacheHit: "X-OmniRoute-Cache-Hit",
  compression: "X-OmniRoute-Compression",
  costSaved: "X-OmniRoute-Cost-Saved",
  • Step 2: Import the parser and declare the capture variable

In open-sse/handlers/chatCore.ts, line 7, add resolveCompressionHeader to the existing import from ./chatCore/headers.ts:

import { getHeaderValueCaseInsensitive, isNoMemoryRequested, resolveCompressionHeader } from "./chatCore/headers.ts";

Near the other compression-scoped lets (~line 1321, beside let cavemanOutputModeApplied = false;), add:

  let compressionResponseMeta: string | null = null;
  • Step 3: Parse the header and add name keys to the combos map

In the compression block, replace the namedCombos build (~line 1373) so it keys by both id and lowercased name, and parse the header right after:

      let namedCombos: Record<string, CompressionPipelineStep[]> = {};
      try {
        const { listCompressionCombos } = await import("../../src/lib/db/compressionCombos.ts");
        namedCombos = Object.fromEntries(
          listCompressionCombos().flatMap((c) => [
            [c.id, c.pipeline],
            [c.name.toLowerCase(), c.pipeline],
          ])
        );
      } catch (err) {
        log?.debug?.(
          "COMPRESSION",
          "Named combos load skipped: " + (err instanceof Error ? err.message : String(err))
        );
      }
      // Phase 3: per-request override. Unknown values fall through in the resolver (never error).
      const compressionHeader = resolveCompressionHeader(clientRawRequest?.headers ?? null);
      if (compressionHeader) {
        log?.debug?.("COMPRESSION", `x-omniroute-compression header: ${compressionHeader}`);
      }

(The active-profile lookup keys on config.activeComboId, always a UUID/slug id, so the added name keys are inert for it — one map serves both paths. A combo named off/default cannot be selected by name because the keyword branches run first; note this in the docs.)

  • Step 4: Thread the header into both selector calls

Append compressionHeader as the last argument to selectCompressionStrategy (~line 1532) and selectCompressionPlan (~line 1601):

      const modeBeforeOutputTransform = selectCompressionStrategy(
        config,
        compressionComboKey,
        estimatedTokens,
        body as Record<string, unknown>,
        { provider, targetFormat, model: effectiveModel },
        namedCombos,
        compressionHeader
      );
      const compressionPlan = selectCompressionPlan(
        config,
        compressionComboKey,
        estimatedTokens,
        compressionInputBody,
        { provider, targetFormat, model: effectiveModel },
        namedCombos,
        compressionHeader
      );
  • Step 5: Capture the resolved meta

Immediately after const mode = compressionPlan.mode as CompressionConfig["defaultMode"]; (~line 1609), add (importing formatCompressionMeta alongside the other strategySelector imports destructured at ~line 1349):

      compressionResponseMeta = formatCompressionMeta(compressionPlan);

And add formatCompressionMeta to the destructured import from ../services/compression/strategySelector.ts (~line 1349):

        formatCompressionMeta,
  • Step 6: Emit the response header at both build sites

In the non-streaming JSON path, right after the attachOmniRouteMetaHeaders(responseHeaders, { ... }); call (~line 4582), add:

    if (compressionResponseMeta) {
      responseHeaders[OMNIROUTE_RESPONSE_HEADERS.compression] = compressionResponseMeta;
    }

In the streaming path, right after the responseHeaders object literal closes (the }; after "x-omniroute-request-id": pendingRequestId,, ~line 4708), add:

  if (compressionResponseMeta) {
    responseHeaders[OMNIROUTE_RESPONSE_HEADERS.compression] = compressionResponseMeta;
  }
  • Step 7: Typecheck + cycle/source guard

Run: npm run typecheck:core Expected: exit 0.

Run: npm run check:cycles Expected: exit 0 — the resolver still has no src/lib/db import (header logic lives in strategySelector.ts, name-key map built in chatCore).

  • Step 8: Run the compression suite again (no regression from wiring)

Run: node --import tsx --import ./open-sse/utils/setupPolyfill.ts --import ./tests/_setup/isolateDataDir.ts --test --test-force-exit "tests/unit/compression/**/*.test.ts" Expected: PASS.

  • Step 9: Commit
git add src/shared/constants/headers.ts open-sse/handlers/chatCore.ts
git commit -m "feat(compression): wire x-omniroute-compression header + response header (Phase 3)"

Task 4: Docs + file-size + full validation

Files:

  • Modify: docs/reference/API_REFERENCE.md

  • Modify: docs/compression/COMPRESSION_GUIDE.md

  • Modify (if needed): config/quality/file-size-baseline.json

  • Step 1: Document the header (API reference)

In docs/reference/API_REFERENCE.md, near the x-omniroute-no-memory documentation, add a section:

### `x-omniroute-compression`

Per-request override of the compression plan. Highest precedence — beats the routing-combo
override, the active profile, auto-trigger, and the panel Default. Values:

| Value | Effect |
|-------|--------|
| `off` | No compression for this request. |
| `default` | The panel-derived Default profile (ignores the active profile). |
| `engine:<id>` | A single engine when enabled, e.g. `engine:rtk`. |
| `<combo>` | A named combo, matched by name (case-insensitive) first, then by id. |

Unknown values are ignored (the request is never rejected). The applied plan is echoed in the
`X-OmniRoute-Compression: <mode>; source=<source>` response header. The master compression
switch is a hard gate: when compression is disabled globally, this header cannot enable it.
  • Step 2: Document the header (compression guide)

In docs/compression/COMPRESSION_GUIDE.md, add a short "Per-request override" subsection mirroring the table above (one paragraph + the value table). Keep all angle-bracket tokens (engine:<id>, <combo>, <mode>, <source>) inside backticks or fenced blocks (the docs are MDX-compiled).

  • Step 3: Validate MDX

Run: npx fumadocs-mdx Expected: [MDX] generated files in ...ms with no error.

  • Step 4: Reconcile the file-size baseline if chatCore crossed its pin

Run: npm run check:file-size Expected: PASS. If it reports chatCore.ts: <actual> > <frozen>, update config/quality/file-size-baseline.json: set the open-sse/handlers/chatCore.ts frozen value to the reported <actual> and add a justification key:

  "_rebaseline_2026_06_21_phase3_request_header": "Compression Phase 3 (x-omniroute-compression per-request header) own growth: chatCore.ts <frozen>-><actual> at the existing compression-dispatch chokepoint — parse the header (resolveCompressionHeader), add lowercased-name keys to the namedCombos map, thread it as the new last arg to selectCompressionStrategy + selectCompressionPlan, capture formatCompressionMeta(compressionPlan) into compressionResponseMeta, and emit X-OmniRoute-Compression at the two response-header build sites. The resolver stays pure (planFromHeader + source live in open-sse/services/compression/strategySelector.ts, <cap). Cohesive wiring at the existing chokepoint, mirroring the Phase 2 rebaseline; not extractable without hiding the dispatch boundary. Structural shrink of chatCore.ts tracked in #3501. Covered by tests/unit/compression/compression-header-dispatch.test.ts + tests/unit/chatcore-headers.test.ts.",
  • Step 5: Run the full validation suite

Run each and confirm exit 0 / green:

  • npm run typecheck:core
  • npm run lint
  • npm run check:cycles
  • npm run check:file-size
  • node --import tsx --import ./open-sse/utils/setupPolyfill.ts --import ./tests/_setup/isolateDataDir.ts --test --test-force-exit "tests/unit/compression/**/*.test.ts" (compression subset) and tests/unit/chatcore-headers.test.ts
  • npm run test:unit (full unit suite — confirms no cross-suite regression)

Expected: all green. Fix any red before proceeding (do not weaken assertions to pass).

  • Step 6: Commit
git add docs/reference/API_REFERENCE.md docs/compression/COMPRESSION_GUIDE.md config/quality/file-size-baseline.json
git commit -m "docs(compression): document x-omniroute-compression header + file-size reconcile (Phase 3)"

Self-Review (run by the author after writing)

1. Spec coverage

  • §2 contract (off/default/engine/combo, unknown→ignore) → Task 1 planFromHeader + tests. ✓
  • Decision A (name-first combo) → Task 1 planFromHeader combos[lower] ?? combos[h] + Task 3 name-key map + test. ✓
  • Decision B (bypass auto-trigger) → Task 1 header-first in resolveBasePlan + test. ✓
  • §3.1 precedence (header at top) → Task 1 + precedence tests (beats active profile, routing override). ✓
  • §3.2 source on DerivedPlan → Task 1 CompressionSource + withSource + tests. ✓
  • §3.3 parser → Task 2. ✓
  • §3.4 threading + id+name map → Task 1 (signatures) + Task 3 (chatCore). ✓
  • §4 response header → Task 3 (constant + two sites) + formatCompressionMeta test. ✓
  • §5 error handling (unknown→fall-through; disabled engine→fall-through; unconditional gating; master-off boundary) → Task 1 tests. ✓
  • §7 testing (parser, resolver, precedence, source guard) → Tasks 1, 2; source guard via check:cycles in Task 3/4. ✓
  • §8 scope (no bare modes, no UI, chat-only) → respected; planFromHeader rejects bare mode names (they fall through), no UI/non-chat files touched. ✓

2. Placeholder scan: none — every code step shows full code; commands have expected output.

3. Type consistency: DerivedPlan/CompressionSource defined in Task 1 Step 3, imported in Step 5; planFromHeader(config, header, combos) / formatCompressionMeta(plan) signatures identical across Task 1 tests, implementation, and Task 3 usage; resolveCompressionHeader(headers) identical in Task 2 and Task 3; header appended as the same final positional arg in getEffectiveMode/selectCompressionPlan/selectCompressionStrategy. ✓

Note for the implementer: typecheck:core, check:cycles, check:file-size, and lint are existing package.json scripts. There is no test:compression script — run the compression subset directly with the node test runner (node --import tsx --import ./open-sse/utils/setupPolyfill.ts --import ./tests/_setup/isolateDataDir.ts --test --test-force-exit "tests/unit/compression/**/*.test.ts"), or run the whole unit suite with npm run test:unit. Confirm any other script name in package.json before running.