betahi-copilot-bridge 0.1.8 → 0.20.0

package/README.md

@@ -1,7 +1,7 @@
 <h1 align="center">copilot-bridge</h1>
 
 <p align="center">
-  <a href="https://www.npmjs.com/package/betahi-copilot-bridge"><img src="https://img.shields.io/npm/v/betahi-copilot-bridge.svg?v=0.1.8" alt="npm version"></a>
+  <a href="https://www.npmjs.com/package/betahi-copilot-bridge"><img src="https://img.shields.io/npm/v/betahi-copilot-bridge.svg?v=0.20.0" alt="npm version"></a>
   <a href="https://github.com/betahi/copilot-bridge/blob/main/LICENSE"><img src="https://img.shields.io/npm/l/betahi-copilot-bridge.svg" alt="license"></a>
 </p>
 
@@ -83,6 +83,7 @@ this block; the bridge regenerates it on every start.
 ```toml
 # >>> copilot-bridge managed block — auto-generated, do not edit between markers >>>
 model_provider = "bridge"
+model_supports_reasoning_summaries = true
 
 [model_providers.bridge]
 name = "Copilot Bridge"
@@ -102,11 +103,6 @@ model = "gpt-5.3-codex"
 model_reasoning_effort = "high"
 ```
 
-If `model_reasoning_effort` is omitted, the bridge leaves it unset and does not
-send a reasoning effort for Codex requests (a startup info log also reminds you).
-
-`model_reasoning_effort` only affects Codex requests; it does not affect Claude.
-
 That's it — `codex exec '...'` will now route through the bridge to Copilot.
 
 Use `--no-codex-setup` to skip the managed-block writer entirely (e.g. if you
@@ -194,16 +190,23 @@ accepts upstream.
 
 ### Claude family — translated to chat completions
 
-| Model                | Reasoning efforts                       | Notes                                  |
-| -------------------- | --------------------------------------- | -------------------------------------- |
-| `claude-opus-4.7`    | `low`, `medium`, `high`, `xhigh`, `max` | Effort sent as `output_config.effort`. |
-| `claude-opus-4.6`    | `low`, `medium`, `high`                 |                                        |
-| `claude-opus-4.6-1m` | `low`, `medium`, `high`                 | 1M-token context window.               |
-| `claude-sonnet-4.6`  | `low`, `medium`, `high`                 |                                        |
-| `claude-opus-4.5`    | —                                       | Reasoning not accepted upstream.       |
-| `claude-sonnet-4.5`  | —                                       | Reasoning not accepted upstream.       |
-| `claude-sonnet-4`    | —                                       | Reasoning not accepted upstream.       |
-| `claude-haiku-4.5`   | —                                       | Reasoning not accepted upstream.       |
+| Model                            | Reasoning efforts                       | Notes                                  |
+| -------------------------------- | --------------------------------------- | -------------------------------------- |
+| `claude-opus-4.7`                | `medium`                                | Effort sent as `output_config.effort`. |
+| `claude-opus-4.7-1m`             | `low`, `medium`, `high`, `xhigh`        | 1M-token context window; effort sent as `output_config.effort`. |
+| `claude-opus-4.7-high`           | `high`                                  | Fixed high reasoning; effort sent as `output_config.effort`. |
+| `claude-opus-4.7-xhigh`          | `xhigh`                                 | Fixed extra-high reasoning; effort sent as `output_config.effort`. |
+| `claude-opus-4.6`                | `low`, `medium`, `high`                 |                                        |
+| `claude-opus-4.6-1m`             | `low`, `medium`, `high`                 | 1M-token context window.               |
+| `claude-sonnet-4.6`              | `low`, `medium`, `high`                 |                                        |
+| `claude-opus-4.5`                | —                                       | Reasoning not accepted upstream.       |
+| `claude-sonnet-4.5`              | —                                       | Reasoning not accepted upstream.       |
+| `claude-sonnet-4`                | —                                       | Reasoning not accepted upstream.       |
+| `claude-haiku-4.5`               | —                                       | Reasoning not accepted upstream.       |
+
+For Claude Opus 4.7, both Codex CLI and Claude Code can use
+`claude-opus-4.7` with reasoning effort `high` or `xhigh`; the bridge routes the
+request to the matching upstream reasoning variant.
 
 ### Gemini family — translated to chat completions
 

package/dist/main.js

@@ -439,6 +439,18 @@ const BEGIN_MARK = "# >>> copilot-bridge managed block — auto-generated, do no
 const END_MARK = "# <<< copilot-bridge managed block — edits outside this block are preserved <<<";
 const LEGACY_BEGIN_MARK = "# >>> copilot-bridge managed (do not edit) >>>";
 const LEGACY_END_MARK = "# <<< copilot-bridge managed (do not edit) <<<";
+function normalizeCodexConfigReasoningEffort(value) {
+	switch (value?.toLowerCase()) {
+		case "none":
+		case "minimal":
+		case "low":
+		case "medium":
+		case "high":
+		case "xhigh": return value.toLowerCase();
+		case "max": return "xhigh";
+		default: return;
+	}
+}
 function tomlEscape(value) {
 	return value.replace(/\\/g, "\\\\").replace(/"/g, "\\\"");
 }
@@ -447,6 +459,7 @@ function buildManagedBlock(input) {
 	const lines = [];
 	lines.push(BEGIN_MARK);
 	if (settings.setAsDefault) lines.push(`model_provider = "${tomlEscape(settings.providerId)}"`);
+	lines.push("model_supports_reasoning_summaries = true");
 	lines.push("");
 	lines.push(`[model_providers.${settings.providerId}]`);
 	lines.push(`name = "${tomlEscape(settings.providerName)}"`);
@@ -502,11 +515,25 @@ function setTopScalar(topSection, key, value) {
 	if (topSection.length === 0) return `${line}\n`;
 	return `${line}\n${topSection.startsWith("\n") ? "" : ""}${topSection}`;
 }
+function removeTopKey(topSection, key) {
+	const lines = topSection.split("\n");
+	const re = /* @__PURE__ */ new RegExp(`^\\s*${key}\\s*=`);
+	return lines.filter((line) => !re.test(line)).join("\n");
+}
+function sanitizeTopReasoningEffort(topSection) {
+	const match = topSection.match(scalarRegex("model_reasoning_effort"));
+	if (!match) return topSection;
+	const normalized = normalizeCodexConfigReasoningEffort(match[1]);
+	if (!normalized) return removeTopKey(topSection, "model_reasoning_effort");
+	if (normalized === match[1]) return topSection;
+	return topSection.replace(scalarRegex("model_reasoning_effort"), `model_reasoning_effort = "${tomlEscape(normalized)}"`);
+}
 function applyUserScalars(content, input) {
 	const { top, rest } = splitTopSection(content);
-	let nextTop = top;
+	let nextTop = removeTopKey(top, "model_supports_reasoning_summaries");
+	nextTop = sanitizeTopReasoningEffort(nextTop);
 	nextTop = setTopScalar(nextTop, "model", input.model);
-	nextTop = setTopScalar(nextTop, "model_reasoning_effort", input.modelReasoningEffort);
+	nextTop = setTopScalar(nextTop, "model_reasoning_effort", normalizeCodexConfigReasoningEffort(input.modelReasoningEffort));
 	if (nextTop === top) return content;
 	if (rest.length === 0) return nextTop.endsWith("\n") ? nextTop : `${nextTop}\n`;
 	const sep = nextTop.endsWith("\n") ? "" : "\n";
@@ -649,21 +676,43 @@ const MODEL_CAPABILITIES = [
 			default: "medium"
 		}
 	},
-	{
-		id: "claude-opus-4.7",
-		fallback: "chat-completions",
-		reasoningField: "output_config.effort",
-		reasoning: {
+	...[
+		{
+			id: "claude-opus-4.7",
+			supported: ["medium"],
+			defaultEffort: "medium"
+		},
+		{
+			id: "claude-opus-4.7-1m-internal",
+			aliases: ["claude-opus-4.7-1m"],
 			supported: [
 				"low",
 				"medium",
 				"high",
-				"xhigh",
-				"max"
+				"xhigh"
 			],
-			default: "medium"
+			defaultEffort: "medium"
+		},
+		{
+			id: "claude-opus-4.7-high",
+			supported: ["high"],
+			defaultEffort: "high"
+		},
+		{
+			id: "claude-opus-4.7-xhigh",
+			supported: ["xhigh"],
+			defaultEffort: "xhigh"
 		}
-	},
+	].map(({ id, aliases, supported, defaultEffort }) => ({
+		id,
+		aliases,
+		fallback: "chat-completions",
+		reasoningField: "output_config.effort",
+		reasoning: {
+			supported,
+			default: defaultEffort
+		}
+	})),
 	{
 		id: "claude-opus-4.6",
 		fallback: "chat-completions",
@@ -1155,7 +1204,7 @@ const normalizeClaudeOpus47Effort = (value) => {
 		default: return;
 	}
 };
-const getEnvValueCaseInsensitive = (env, key) => {
+const getEnvValueCaseInsensitive$1 = (env, key) => {
 	const direct = env[key];
 	if (typeof direct === "string") return direct;
 	const lower = key.toLowerCase();
@@ -1163,7 +1212,7 @@ const getEnvValueCaseInsensitive = (env, key) => {
 };
 const getConfiguredReasoningEffort = (client, claudeSettingsEnv) => {
 	if (client !== "claude") return;
-	return process.env.MODEL_REASONING_EFFORT ?? getEnvValueCaseInsensitive(claudeSettingsEnv, "MODEL_REASONING_EFFORT");
+	return process.env.MODEL_REASONING_EFFORT ?? getEnvValueCaseInsensitive$1(claudeSettingsEnv, "MODEL_REASONING_EFFORT");
 };
 const getRequestedReasoningEffort = (payload, claudeSettingsEnv, client) => {
 	const configuredReasoningEffort = getConfiguredReasoningEffort(client, claudeSettingsEnv);
@@ -1173,7 +1222,9 @@ const getRequestedReasoningEffort = (payload, claudeSettingsEnv, client) => {
 const getRequestedClaudeOpus47Effort = (payload, claudeSettingsEnv, client) => {
 	if (!isClaudeOpus47Model$1(payload.model)) return;
 	const configuredReasoningEffort = getConfiguredReasoningEffort(client, claudeSettingsEnv);
-	return payload.output_config?.effort ?? normalizeClaudeOpus47Effort(payload.reasoning_effort) ?? normalizeClaudeOpus47Effort(configuredReasoningEffort);
+	const requested = normalizeClaudeOpus47Effort(payload.reasoning_effort) ?? normalizeClaudeOpus47Effort(configuredReasoningEffort) ?? payload.output_config?.effort;
+	if (!requested) return;
+	return clampReasoningEffort(payload.model, requested)?.effort;
 };
 const sanitizeUserIdentifier = (user) => {
 	if (!user) return;
@@ -1366,20 +1417,21 @@ embeddingRoutes.post("/", async (c) => {
 const normalizeModelId = (modelId) => modelId.trim().toLowerCase().replaceAll(/[\s._-]+/g, "");
 const stripSnapshotSuffix = (modelId) => {
 	const id = modelId.trim().toLowerCase().replace(/\[1m\]$/, "");
-	const claudeSnapshotMatch = id.match(/^claude-(opus|sonnet|haiku)-(\d)-(\d)(-1m)?(?:-\d{8})?$/);
+	if (id === "claude-opus-4.7-1m") return "claude-opus-4.7-1m-internal";
+	const claudeSnapshotMatch = id.match(/^claude-(opus|sonnet|haiku)-(\d)-(\d)((?:-[a-z0-9]+)*?)(?:-\d{8})?$/);
 	if (claudeSnapshotMatch) {
-		const [, family, major, minor, contextWindow = ""] = claudeSnapshotMatch;
-		return `claude-${family}-${major}.${minor}${contextWindow}`;
+		const [, family, major, minor, suffix = ""] = claudeSnapshotMatch;
+		return `claude-${family}-${major}.${minor}${suffix}`;
 	}
 	return id;
 };
 const getAliasCandidates = (modelId) => {
 	const canonicalModelId = stripSnapshotSuffix(modelId);
 	const aliases = new Set([canonicalModelId]);
-	const canonicalClaudeMatch = canonicalModelId.match(/^claude-(opus|sonnet|haiku)-(\d+)\.(\d+)(-1m)?$/);
+	const canonicalClaudeMatch = canonicalModelId.match(/^claude-(opus|sonnet|haiku)-(\d+)\.(\d+)((?:-[a-z0-9]+)*)$/);
 	if (canonicalClaudeMatch) {
-		const [, family, major, minor, contextWindow = ""] = canonicalClaudeMatch;
-		aliases.add(`claude-${family}-${major}-${minor}${contextWindow}`);
+		const [, family, major, minor, suffix = ""] = canonicalClaudeMatch;
+		aliases.add(`claude-${family}-${major}-${minor}${suffix}`);
 		return [...aliases];
 	}
 	const familyMatch = canonicalModelId.match(/^[a-z]+(?:-[a-z]+)*-\d+(?:\.\d+)?/);
@@ -1526,12 +1578,42 @@ const normalizeClaudeModelAlias = (model) => {
 	const trimmed = model.trim().toLowerCase();
 	if (trimmed === "opus[1m]") return "claude-opus-4.6-1m";
 	const normalized = trimmed.replace(/\[1m\]$/, "");
-	if (normalized === "opus") return "claude-opus";
-	if (normalized === "sonnet") return "claude-sonnet";
-	if (normalized === "haiku") return "claude-haiku";
-	if (/^claude-(opus|sonnet|haiku)-\d-\d(?:-1m)?$/.test(normalized)) return normalized.replace(/^claude-(opus|sonnet|haiku)-(\d)-(\d)(-1m)?$/, "claude-$1-$2.$3$4");
-	if (/^claude-opus-4-7-\d{8}$/.test(normalized)) return "claude-opus-4.7";
-	return normalized;
+	const prefixed = normalized.startsWith("opus-") ? `claude-${normalized}` : normalized;
+	if (prefixed === "opus") return "claude-opus";
+	if (prefixed === "sonnet") return "claude-sonnet";
+	if (prefixed === "haiku") return "claude-haiku";
+	const claudeSnapshotMatch = prefixed.match(/^claude-(opus|sonnet|haiku)-(\d)-(\d)((?:-[a-z0-9]+)*?)(?:-\d{8})?$/);
+	if (claudeSnapshotMatch) {
+		const [, family, major, minor, suffix = ""] = claudeSnapshotMatch;
+		return `claude-${family}-${major}.${minor}${suffix}`;
+	}
+	return prefixed;
+};
+function normalizeClaudeReasoningEffortForRouting(value) {
+	switch (value?.toLowerCase()) {
+		case "low":
+		case "medium":
+		case "high":
+		case "xhigh":
+		case "max": return value.toLowerCase();
+		default: return;
+	}
+}
+const getEnvValueCaseInsensitive = (env, key) => {
+	const direct = env[key];
+	if (typeof direct === "string") return direct;
+	const lower = key.toLowerCase();
+	return Object.entries(env).find(([k]) => k.toLowerCase() === lower)?.[1];
+};
+const getConfiguredClaudeReasoningEffort = (settings) => process.env.MODEL_REASONING_EFFORT ?? getEnvValueCaseInsensitive(settings?.env ?? {}, "MODEL_REASONING_EFFORT");
+const routeClaudeOpus47ByEffort = (model, requestedEffort) => {
+	if (model !== "claude-opus-4.7") return model;
+	switch (normalizeClaudeReasoningEffortForRouting(requestedEffort)) {
+		case "high": return "claude-opus-4.7-high";
+		case "xhigh":
+		case "max": return "claude-opus-4.7-xhigh";
+		default: return model;
+	}
 };
 const getConfiguredClaudeDefaultModel = (settings) => {
 	if (!settings) return;
@@ -1551,8 +1633,8 @@ const resolveClaudeRequestedModel = (model, settings) => {
 	if (model.trim().toLowerCase() !== "definitely-not-a-real-model") return model;
 	return getConfiguredClaudeDefaultModel(settings) ?? model;
 };
-function translateModelName(model, settings) {
-	return resolveUpstreamModelId(normalizeClaudeModelAlias(resolveClaudeRequestedModel(model, settings)));
+function translateModelName(model, settings, requestedReasoningEffort) {
+	return resolveUpstreamModelId(routeClaudeOpus47ByEffort(normalizeClaudeModelAlias(resolveClaudeRequestedModel(model, settings)), requestedReasoningEffort ?? getConfiguredClaudeReasoningEffort(settings)));
 }
 function isClaudeModel(modelId) {
 	return modelId.startsWith("claude-");
@@ -1582,12 +1664,12 @@ function getClaudeOpus47Effort(payload) {
 	return "xhigh";
 }
 function translateThinking(payload, settings) {
-	if (!isClaudeOpus47Model(translateModelName(payload.model, settings))) return;
+	if (!isClaudeOpus47Model(translateModelName(payload.model, settings, payload.reasoning_effort))) return;
 	if (payload.thinking?.type === "adaptive") return { type: "adaptive" };
 	return payload.thinking?.type === "enabled" ? { type: "adaptive" } : void 0;
 }
 function translateOutputConfig(payload, settings) {
-	if (!isClaudeOpus47Model(translateModelName(payload.model, settings))) return;
+	if (!isClaudeOpus47Model(translateModelName(payload.model, settings, payload.reasoning_effort))) return;
 	const effort = getClaudeOpus47Effort(payload);
 	return effort ? { effort } : void 0;
 }
@@ -1603,7 +1685,7 @@ function normalizeReasoningEffort(value) {
 	}
 }
 function translateReasoningEffort(payload, settings) {
-	const modelId = translateModelName(payload.model, settings);
+	const modelId = translateModelName(payload.model, settings, payload.reasoning_effort);
 	if (isClaudeOpus47Model(modelId)) return;
 	if (payload.reasoning_effort) return sanitizeReasoningEffortForModel(modelId, normalizeReasoningEffort(payload.reasoning_effort));
 	if (isClaudeModel(modelId)) return;
@@ -1616,7 +1698,7 @@ function translateReasoningEffort(payload, settings) {
 	return sanitizeReasoningEffortForModel(modelId, "xhigh");
 }
 function translateToOpenAI(payload, settings, toolNameMapper) {
-	const model = translateModelName(payload.model, settings);
+	const model = translateModelName(payload.model, settings, payload.reasoning_effort);
 	const mapper = toolNameMapper ?? createAnthropicToolNameMapper(payload.tools, { ...getToolNameMapperOptionsForModel(model) });
 	const tools = translateAnthropicToolsToOpenAI(payload.tools, mapper);
 	return {
@@ -2532,9 +2614,19 @@ const removeReasoningEffort = (reasoning) => {
 	return Object.keys(next).length > 0 ? next : void 0;
 };
 const isPlainObject = (value) => typeof value === "object" && value !== null && !Array.isArray(value);
-const normalizeCodexResponsesRequest = (payload) => {
+const routeClaudeOpus47ByReasoningEffort = (model, effort) => {
+	if (model !== "claude-opus-4.7") return model;
+	switch (typeof effort === "string" ? effort.toLowerCase() : void 0) {
+		case "high": return "claude-opus-4.7-high";
+		case "xhigh":
+		case "max": return "claude-opus-4.7-xhigh";
+		default: return model;
+	}
+};
+const normalizeCodexResponsesRequest = (payload, configuredReasoningEffort) => {
 	const parsed = codexResponsesRequestSchema.parse(payload);
-	const canonical = resolveModelId(parsed.model);
+	const requestedReasoningEffort = (isPlainObject(parsed.reasoning) ? parsed.reasoning : void 0)?.effort ?? configuredReasoningEffort;
+	const canonical = routeClaudeOpus47ByReasoningEffort(resolveModelId(parsed.model), requestedReasoningEffort);
 	const capability = getModelCapability(canonical);
 	if (!capability) return parsed;
 	const next = {
@@ -2545,21 +2637,30 @@ const normalizeCodexResponsesRequest = (payload) => {
 		if ("reasoning" in next) delete next.reasoning;
 		return next;
 	}
-	if ("reasoning" in next) if (!isPlainObject(next.reasoning)) delete next.reasoning;
+	if ("reasoning" in next) if (!isPlainObject(next.reasoning)) if (configuredReasoningEffort === void 0 || configuredReasoningEffort === null) delete next.reasoning;
+	else {
+		const clamped = clampReasoningEffort(canonical, configuredReasoningEffort);
+		if (clamped) next.reasoning = { effort: clamped.effort };
+	}
 	else {
 		const incoming = next.reasoning;
-		if (incoming.effort === void 0 || incoming.effort === null) {
+		const effort = incoming.effort ?? configuredReasoningEffort;
+		if (effort === void 0 || effort === null) {
 			const reasoning = removeReasoningEffort(incoming);
 			if (reasoning) next.reasoning = reasoning;
 			else delete next.reasoning;
 		} else {
-			const clamped = clampReasoningEffort(canonical, incoming.effort);
+			const clamped = clampReasoningEffort(canonical, effort);
 			if (clamped) next.reasoning = {
 				...incoming,
 				effort: clamped.effort
 			};
 		}
 	}
+	else if (configuredReasoningEffort !== void 0 && configuredReasoningEffort !== null) {
+		const clamped = clampReasoningEffort(canonical, configuredReasoningEffort);
+		if (clamped) next.reasoning = { effort: clamped.effort };
+	}
 	const text = isPlainObject(next.text) ? next.text : void 0;
 	if (text && capability.textVerbosity && "verbosity" in text) {
 		const { supported, default: defaultVerbosity } = capability.textVerbosity;
@@ -2597,6 +2698,9 @@ const logResponsesUpstreamError = async (message, response, context) => {
 		request: JSON.stringify(summarizeResponsesRequestForDiagnostics(context.request))
 	});
 };
+const readConfiguredCodexReasoningEffort = async () => {
+	return normalizeCodexConfigReasoningEffort((await readCodexUserConfigFromDisk(process.env.CODEX_CONFIG_PATH ?? CODEX_DEFAULTS.configPath)).modelReasoningEffort);
+};
 responsesRoutes.post("/", async (c) => {
 	try {
 		await checkRateLimit();
@@ -2604,7 +2708,7 @@ responsesRoutes.post("/", async (c) => {
 		if (error instanceof RateLimitError) return c.json({ error: { message: error.message } }, 429);
 		throw error;
 	}
-	const payload = normalizeCodexResponsesRequest(await c.req.json());
+	const payload = normalizeCodexResponsesRequest(await c.req.json(), await readConfiguredCodexReasoningEffort());
 	const provider = getCopilotProviderContext(c.get("config"));
 	const search = new URL(c.req.url).search;
 	const capability = getModelCapability(payload.model);
@@ -2758,6 +2862,8 @@ const startServer = (config) => serve({
 
 //#endregion
 //#region src/start.ts
+const getPublicModelId = (id) => id === "claude-opus-4.7-1m-internal" ? "claude-opus-4.7-1m" : id;
+const unique = (ids) => [...new Set(ids)];
 const fetchAvailableModels = async (config) => {
 	try {
 		const response = await fetchCopilot(getCopilotProviderContext(config), "/models", {
@@ -2897,10 +3003,13 @@ const start = defineCommand({
 			consola.warn(`Could not update codex config (${codexConfigPath}):`, error);
 		}
 		const models = await fetchAvailableModels(config);
-		const supportedIds = new Set(MODEL_CAPABILITIES.map((m) => m.id));
-		const pickable = models.length > 0 ? models.filter((id) => supportedIds.has(id)) : MODEL_CAPABILITIES.map((m) => m.id);
-		const finalPickable = pickable.length > 0 ? pickable : MODEL_CAPABILITIES.map((m) => m.id);
-		if (models.length > 0) consola.info(`Available models:\n${models.map((id) => `- ${id}${supportedIds.has(id) ? " (bridge-supported)" : ""}`).join("\n")}`);
+		const supportedIds = new Set(MODEL_CAPABILITIES.flatMap((m) => [m.id, ...m.aliases ?? []]));
+		const fallbackModelIds = unique(MODEL_CAPABILITIES.map((m) => getPublicModelId(m.id)));
+		const pickable = models.length > 0 ? unique(models.filter((id) => supportedIds.has(id)).map((id) => getPublicModelId(id))) : fallbackModelIds;
+		const finalPickable = pickable.length > 0 ? pickable : fallbackModelIds;
+		if (models.length > 0) consola.info(`Available models:\n${models.map((id) => {
+			return `- ${getPublicModelId(id)}${supportedIds.has(id) ? " (bridge-supported)" : ""}`;
+		}).join("\n")}`);
 		else consola.warn("Could not fetch model list from upstream Copilot API");
 		consola.box([
 			`🌐 Usage viewer`,