feat(copilot): per-model GitHub Copilot context tier (carry #3282)

docs-site/src/content/docs/guides/model-routing.md

@@ -105,6 +105,15 @@ Routing and catalog visibility are separate controls:
   switching it on restores that selection. A remembered selection never applies a limit while disabled.
   Sending `{ "setAll": true }` without `value` enables all configured providers at the current
   global value and replaces their remembered selections.
+- A GitHub Copilot model that supports the upstream long-context tier can opt in with
+  `providers.github-copilot.modelContextTiers.<model> = "long_context"` or
+  `ocx provider edit github-copilot --model-context-tier <model>=long_context`.
+  The tier sends `contextTier: "long_context"` on Copilot Chat and Responses requests.
+  It raises the advertised window only when that exact model has a known per-model window
+  in `modelContextWindows` (for example, `1000000`); an unknown model keeps its live window.
+  An explicit per-model window wins, then the existing provider cap applies.
+  The `"default"` tier retains normal live context metadata.
+  The dashboard control is planned separately.
 
 ```json
 {

docs-site/src/content/docs/reference/cli/providers-accounts.md

@@ -16,7 +16,7 @@ both `--adapter` and `--base-url`.
 | --- | --- | --- |
 | `list` | `--json`, `--jsonl` | List configured providers and the remaining registry entries; `--jsonl` emits one configured provider object per line. |
 | `add <name>` | `--adapter <adapter>`, `--base-url <url>`, `--api-key <key>`, `--default-model <model>`, `--set-default`, `--force`, `--json`, `--sync` | Add a registry/custom provider. `--force` overwrites; `--sync` refreshes a running proxy in human-output mode. |
-| `edit <name>` | provider field flags, `--headers <json>`, `--json` | Edit validated live provider fields without replacing key pools. `--headers` merges custom request headers; pass `{}` or `-` to clear them. |
+| `edit <name>` | provider field flags, `--headers <json>`, `--model-context-tier <model=default\|long_context>`, `--json` | Edit validated live provider fields without replacing key pools. Repeat `--model-context-tier` for multiple Copilot models. `--headers` merges custom request headers; pass `{}` or `-` to clear them. |
 | `test <name>` | `--json` | Probe the real upstream model endpoint. |
 | `show <name>` | `--json` | Show config with API keys masked. |
 | `remove <name>` | `--json` | Remove a non-default provider; the last provider cannot be removed. |
@@ -34,6 +34,7 @@ ocx provider test ark
 ocx provider add anthropic --api-key sk-ant-... --set-default --sync
 ocx provider add local-dev --adapter openai-chat --base-url http://localhost:11434/v1
 ocx provider show anthropic --json
+ocx provider edit github-copilot --model-context-tier gpt-5.6-luna=long_context
 ocx models --provider anthropic --json
 ocx models live --provider ark --json
 ```

docs-site/src/content/docs/reference/configuration/providers.md

@@ -225,6 +225,7 @@ Providers can expose a built-in shorthand, such as `agy` for `google-antigravity
 | `modelDisplayNames?` | `Record<string, string>` | Durable labels used only for display, keyed by this provider's exact upstream model id. Labels win over provider catalog metadata, survive discovery refreshes and provider edits, and never change authentication, adapter behavior, routing, billing, upstream request construction, the routed `provider/model` selector, or the upstream wire model. Keys are exact and case sensitive. Unknown model ids are kept so a temporarily missing model receives its label when it returns. The map accepts at most 2,000 entries, matching the discovery limit. |
 | `contextWindow?` | `number` | Provider-wide context fallback when upstream metadata is absent; otherwise a cap that retains smaller live metadata. The Models dashboard exposes this separately from `providerContextCaps`. |
 | `modelContextWindows?` | `Record<string, number>` | Per-model context fallbacks/caps. These override `contextWindow`: an unknown window uses the configured value, while smaller live metadata remains authoritative. |
+| `modelContextTiers?` | `Record<string, "default" \| "long_context">` | Per-model upstream tier for `github-copilot`. `long_context` sends `contextTier: "long_context"`. Catalog metadata grows only with an exact per-model `modelContextWindows` value that documents the supported window; that value wins before the provider cap. An unknown model keeps its observed window. Unconfigured passthrough requests retain a caller-supplied `contextTier` field on any provider. |
 | `modelInputModalities?` | `Record<string, string[]>` | Per-model input hints such as `["text"]` or `["text", "image"]`. |
 | `modelMaxInputTokens?` | `Record<string, number>` | Positive per-model max input limits used for catalog auto-compaction hints. |
 | `modelAutoCompactTokenLimits?` | `Record<string, number>` | Positive safe-integer per-model soft auto-compaction budgets. Values can only lower the effective 90%-of-context/max-input envelope and are omitted when no authoritative context window is known. For canonical `openai`, keys must be exact supported native model IDs without provider or account-selector prefixes. Provider PATCH merges entries; set a key to `null` to delete it or the whole field to `null` to clear the map. These `null` tombstones are PATCH-only. |

scripts/test-layout/layout.json

@@ -830,6 +830,7 @@
     "gemini-web-search.test.ts": "adapters/google",
     "generic-oauth-failover.test.ts": "oauth",
     "github-copilot-account-origin.test.ts": "providers/github-copilot",
+    "github-copilot-context-tier.test.ts": "providers/github-copilot",
     "github-copilot-oauth.test.ts": "providers/github-copilot",
     "github-copilot-sse-rewrite.test.ts": "providers/github-copilot",
     "github-copilot-stream-contract.test.ts": "providers/github-copilot",

src/adapters/base.ts

@@ -8,6 +8,8 @@ import type { ProviderRequestSlot } from "../providers/request-pacing";
 /** Metadata about the caller's incoming request, for auth-forwarding adapters. */
 export interface IncomingMeta {
   headers: Headers;
+  /** Canonical routed provider identity for provider-specific wire fields. */
+  providerName?: string;
   translatorBudget: TranslatorBudget;
   abortSignal?: AbortSignal;
   /** Lease acquired before the response is committed; the physical send transfers it to its body. */

src/adapters/openai-chat.ts

@@ -13,6 +13,7 @@ import { isDebugEnabled } from "../lib/debug-settings";
 import { openRouterProviderPayload, resolveOpenRouterRouting } from "../providers/openrouter-routing";
 import { resolveVercelGatewayRouting, vercelGatewayProviderPayload } from "../providers/vercel-gateway-routing";
 import { fastPolicyForModel } from "../providers/service-tier";
+import { applyGithubCopilotContextTier } from "../providers/github-copilot-context";
 import { createAdapterTierMetadata, decideTier, type AdapterTierMetadata } from "../providers/fastwire";
 import {
   isTranslatorBudgetExceededError,
@@ -229,7 +230,7 @@ export function createOpenAIChatAdapter(provider: OcxProviderConfig): ProviderAd
         }
         if (parsed.stream) body.stream_options = { include_usage: true };
 
-        const bodyJson = JSON.stringify(body);
+        const bodyJson = JSON.stringify(applyGithubCopilotContextTier(body, provider, parsed.modelId, incoming?.providerName));
         const actualServiceTier = typeof body.service_tier === "string" ? body.service_tier : null;
         const tierLog = createAdapterTierMetadata(
           parsed.options.tierObservation,

src/adapters/openai-chat/passthrough.ts

@@ -6,6 +6,7 @@ import { applyExplicitChatDeveloperRole } from "./developer-role";
 import { openRouterProviderPayload, resolveOpenRouterRouting } from "../../providers/openrouter-routing";
 import { resolveVercelGatewayRouting, vercelGatewayProviderPayload } from "../../providers/vercel-gateway-routing";
 import { fastPolicyForModel } from "../../providers/service-tier";
+import { applyGithubCopilotContextTier } from "../../providers/github-copilot-context";
 import { canonicalFastTierMarker, decideTier, type ResolvedFastPolicy } from "../../providers/fastwire";
 import { debugProviderDiagnostic } from "../../lib/debug";
 import { isDebugEnabled } from "../../lib/debug-settings";
@@ -53,6 +54,7 @@ export function buildOpenAIChatPassthroughRequest(
   stream: boolean,
   fastPolicy: ResolvedFastPolicy = fastPolicyForModel(provider, modelId, undefined, "chat"),
   fastMode?: boolean,
+  providerName?: string,
 ): AdapterRequest {
   const { url, headers, hasCredential } = openAIChatTransport(provider);
 
@@ -153,7 +155,7 @@ export function buildOpenAIChatPassthroughRequest(
     body.stream_options = rawBody.stream_options;
   }
 
-  const bodyJson = JSON.stringify(body);
+  const bodyJson = JSON.stringify(applyGithubCopilotContextTier(body, provider, modelId, providerName));
 
   if (isDebugEnabled()) {
     let host = "upstream";

src/adapters/openai-responses/passthrough.ts

@@ -3,6 +3,7 @@ import { nameRoutedIdentity, repairIdentityInResponsesBody, stripRoutedIdentity
 import { stripBracketedModelSuffix } from "../openai-chat";
 import { normalizeOpenCodeGoAdditionalTools } from "../opencode-go-additional-tools";
 import { isXaiResponsesDestination } from "../../providers/xai-transport";
+import { applyGithubCopilotContextTier } from "../../providers/github-copilot-context";
 import { Buffer } from "node:buffer";
 import type { IncomingMeta, ProviderAdapter } from "../base";
 import { namespacedToolName, type AdapterEvent, type OcxParsedRequest, type OcxProviderConfig, type OcxUsage, type TierDecision } from "../../types";
@@ -538,7 +539,7 @@ export function createResponsesPassthroughAdapter(provider: OcxProviderConfig):
       if (!isCanonicalOpenAiForwardProvider(provider)) {
         validateFinalCustomToolCompatibility(finalBody, provider.supportsResponsesCustomTools);
       }
-      const body = JSON.stringify(finalBody);
+      const body = JSON.stringify(applyGithubCopilotContextTier(finalBody, provider, parsed.modelId, incoming.providerName));
       const releaseBodyObservation = translatorBudget.observeExternallyCapped(
         "passthrough_serialization",
         Buffer.byteLength(body, "utf8"),

src/cli/provider-runtime.ts

@@ -1,4 +1,4 @@
-import { modelCapabilitiesConfigError } from "../config/provider-validation";
+import { contextTierRecordConfigError, modelCapabilitiesConfigError } from "../config/provider-validation";
 import {
   CliUsageError,
   csv,
@@ -42,7 +42,7 @@ const USAGE = `Usage:
       [--headers <json>] [--enabled <on|off>] [--live-models <on|off>]
       [--retain-models <id,id|->] [--model <id> --text-only]
       [--xai-chat <on|off>]
-      [--allow-private-network <on|off>] [--json]
+      [--allow-private-network <on|off>] [--model-context-tier <model=default|long_context>] [--json]
   ocx provider test <name> [--json]
   ocx provider quota [--refresh] [--json]
   ocx provider resets [--limit <n>] [--json]
@@ -73,6 +73,12 @@ async function edit(argv: string[], deps: RuntimeApiDeps): Promise<void> {
   const liveModels = takeBooleanOption(args, "--live-models");
   const allowPrivateNetwork = takeBooleanOption(args, "--allow-private-network");
   const xaiChat = takeBooleanOption(args, "--xai-chat");
+  const contextTierValues: string[] = [];
+  for (;;) {
+    const value = takeOption(args, "--model-context-tier");
+    if (value === undefined) break;
+    contextTierValues.push(value);
+  }
   const textOnly = takeFlag(args, "--text-only");
   const capabilityModel = takeOption(args, "--model");
   rejectArgs(args, USAGE);
@@ -83,6 +89,21 @@ async function edit(argv: string[], deps: RuntimeApiDeps): Promise<void> {
     if (error) throw new CliUsageError(error, USAGE);
     patch.modelCapabilities = declaration;
   }
+  if (contextTierValues.length) {
+    if (name !== "github-copilot") throw new CliUsageError("--model-context-tier is valid only for provider github-copilot", USAGE);
+    const tiers: Record<string, "default" | "long_context"> = Object.create(null);
+    for (const value of contextTierValues) {
+      const separator = value.indexOf("=");
+      const model = value.slice(0, separator);
+      const tier = value.slice(separator + 1);
+      if (separator < 1 || (tier !== "default" && tier !== "long_context"))
+        throw new CliUsageError("--model-context-tier must use model=default or model=long_context", USAGE);
+      tiers[model] = tier;
+    }
+    const error = contextTierRecordConfigError(tiers);
+    if (error) throw new CliUsageError(error, USAGE);
+    patch.modelContextTiers = tiers;
+  }
   if (xaiChat !== undefined) {
     if (name !== "xai") throw new CliUsageError("--xai-chat is valid only for provider xai", USAGE);
     patch.xaiResponsesOptIn = !xaiChat;

src/cli/provider.ts

@@ -245,6 +245,9 @@ async function handleAdd(args: string[]): Promise<void> {
   if (existingProvider?.modelCapabilities !== undefined && provConfig.modelCapabilities === undefined) {
     provConfig.modelCapabilities = structuredClone(existingProvider.modelCapabilities);
   }
+  if (existingProvider?.modelContextTiers !== undefined && provConfig.modelContextTiers === undefined) {
+    provConfig.modelContextTiers = structuredClone(existingProvider.modelContextTiers);
+  }
   if (textOnly) {
     const modelId = capabilityModel ?? defaultModel ?? provConfig.defaultModel;
     if (!modelId) {

src/codex/catalog/gather-capture.ts

@@ -531,6 +531,7 @@ function providerCatalogFingerprint(name: string, prov: OcxProviderConfig): Reco
     defaultModel: prov.defaultModel ?? null,
     ctx: prov.contextWindow ?? null,
     ctxW: prov.modelContextWindows ?? null,
+    ctxTier: prov.modelContextTiers ?? null,
     maxIn: prov.modelMaxInputTokens ?? null,
     maxOut: prov.modelMaxOutputTokens ?? null,
     autoCompact: prov.modelAutoCompactTokenLimits ?? null,

src/codex/catalog/model-hints.ts

@@ -45,6 +45,7 @@ import type { FastPolicyAuthority } from "../../providers/fastwire";
 import { effectiveGoogleMode, getProviderRegistryEntry, providerMatchesRegistryTransport, registryEntryForProviderDestination } from "../../providers/registry";
 import { parseAntigravityAvailableModels, registerAntigravityDiscoveredWireModels } from "../../providers/antigravity-models";
 import { applyProviderContextCap, providerContextCap, resolveUnknownRoutedContextWindow } from "../../providers/context-cap";
+import { githubCopilotCatalogContextWindow } from "../../providers/github-copilot-context";
 import { clampAutoCompactTokenLimit } from "../../providers/auto-compact-budget";
 import { effectiveModelAliases } from "../../providers/default-aliases";
 import { routedSlug, slugEquals, slugEquivalenceKey, slugsEquivalent } from "../../providers/slug-codec";
@@ -310,9 +311,8 @@ export function applyProviderConfigHints(
     ...modelWithoutServiceTier
   } = model;
   // 已发现窗口只允许被配置值压低；缺窗口时，已开的 Context cap 就是实际窗口。
-  const discoveredWindow = typeof model.contextWindow === "number" && model.contextWindow > 0
-    ? model.contextWindow
-    : undefined;
+  const discoveredWindow = githubCopilotCatalogContextWindow(name, prov, model.id,
+    typeof model.contextWindow === "number" && model.contextWindow > 0 ? model.contextWindow : undefined);
   const projectedLimits = clampObservedModelLimits(staticPolicy.model, {
     ...(discoveredWindow !== undefined ? { contextWindow: discoveredWindow } : {}),
     ...(typeof model.maxInputTokens === "number" && model.maxInputTokens > 0 ? { maxInputTokens: model.maxInputTokens } : {}),

src/config/provider-validation.ts

@@ -383,6 +383,19 @@ function capabilityRecord(value: unknown): value is Record<string, unknown> {
     && [Object.prototype, null].includes(Object.getPrototypeOf(value));
 }
 
+/** Strict per-model context-tier values; PATCH accepts null to clear the entire map. */
+export function contextTierRecordConfigError(value: unknown, allowClear = false): string | null {
+  if (value === undefined || (allowClear && value === null)) return null;
+  if (!capabilityRecord(value)) return "modelContextTiers must be a plain object";
+  if (Object.keys(value).length > MODEL_DISCOVERY_MAX_MODELS) return "modelContextTiers has too many models";
+  for (const [id, tier] of Object.entries(value)) {
+    if (!isValidModelDiscoveryModelId(id) || ["__proto__", "prototype", "constructor"].includes(id))
+      return "modelContextTiers keys must be exact non-reserved model ids";
+    if (tier !== "default" && tier !== "long_context") return "modelContextTiers values must be default or long_context";
+  }
+  return null;
+}
+
 /** Strict writes; only PATCH may carry deletion tombstones. Model IDs are exact. */
 export function modelCapabilitiesConfigError(value: unknown, allowTombstones = false): string | null {
   if (value === undefined || (allowTombstones && value === null)) return null;

src/config/schema/leaf-validators.ts

@@ -10,6 +10,7 @@ import {
   normalizeNonBlankStringArray,
   normalizeAutoReviewModelOverrides,
   modelCapabilitiesConfigError,
+  contextTierRecordConfigError,
   mergeModelCapabilities,
 } from "../provider-validation";
 import { isValidCodexAccountNamespaceTarget } from "../../codex/account-namespace-match";
@@ -279,6 +280,10 @@ const providerNoProxySchema = z.unknown().superRefine((value, ctx) => {
  */
 export const providerConfigSchema = z.object({
   modelCapabilities: modelCapabilitiesSchema.optional(),
+  modelContextTiers: z.unknown().superRefine((value, ctx) => {
+    const error = contextTierRecordConfigError(value);
+    if (error) ctx.addIssue({ code: "custom", message: error });
+  }).optional().transform(value => value as OcxProviderConfig["modelContextTiers"]),
   pinnedReasoningEffort: pinnedReasoningEffortSchema.optional(),
   modelPinnedReasoningEfforts: modelPinnedEffortsSchema.optional(),
   // Validated rather than left to passthrough: an unrecognized strategy would otherwise

src/images/loop.ts

@@ -514,6 +514,7 @@ export async function runWithImageBridge(deps: ImageBridgeDeps): Promise<Respons
         deps.onAttemptSend?.();
         void adapter.runTurn(iterParsed, {
           headers: deps.forwardHeaders ? new Headers(deps.forwardHeaders) : new Headers(),
+          ...(deps.incomingMeta.providerName ? { providerName: deps.incomingMeta.providerName } : {}),
           abortSignal: signal,
           translatorBudget,
           pacingSlot,
@@ -591,6 +592,7 @@ export async function runWithImageBridge(deps: ImageBridgeDeps): Promise<Respons
         } else {
           request = await requestAdapter.buildRequest(iterParsed, {
             headers: deps.forwardHeaders ? new Headers(deps.forwardHeaders) : new Headers(),
+            ...(deps.incomingMeta.providerName ? { providerName: deps.incomingMeta.providerName } : {}),
             abortSignal: headerDeadline.signal,
             translatorBudget,
           });

src/oauth/index.ts

@@ -1588,6 +1588,10 @@ export function upsertOAuthProvider(config: OcxConfig, provider: string): void {
   // Login used to rebuild the whole row from the preset, so catalog data refreshed
   // immediately. Keep that timing without overwriting unrelated operator-owned fields.
   applyOAuthPresetCatalog(next, providerConfig);
+  // The per-model Copilot tier is operator intent, not account or preset metadata.
+  if (existing?.modelContextTiers !== undefined) {
+    next.modelContextTiers = structuredClone(existing.modelContextTiers);
+  }
   // The original Command Code seed was an implementation-owned static catalog, not an
   // operator opt-out. Promote that exact legacy shape when OAuth login refreshes the row.
   if (provider === "command-code" && existing && isLegacyCommandCodeStaticCatalog(existing)) {