Merge origin/master (v0.5.91) into gitea/new_feature
Resolve conflicts: - streamingHandler.js: adopt upstreamResponseHeaders while keeping 0-token detail row avoidance - capabilities.js: preserve user-asserted caps and globalThis slots without local caching of catalogSource - AddCustomModelModal.js & providers/[id]/page.js: wire STT transport marker with custom model edits/assertions - models/custom/route.js & aliasRepo.js: persist custom model transport and invalidate user caps - usageRepo.js: key byApiKey live stats by full API key and keep tail in maskApiKey - UsageStats.js: lazy load charts dynamically
This commit is contained in:
@@ -7,7 +7,7 @@ import {
|
||||
} from "../services/oauthCredentialManager.js";
|
||||
import { normalizeResponsesInput } from "../translator/formats/responsesApi.js";
|
||||
import { fetchImageAsBase64 } from "../translator/concerns/image.js";
|
||||
import { getModelUpstreamId } from "../config/providerModels.js";
|
||||
import { getModelUpstreamId, getProviderModels } from "../config/providerModels.js";
|
||||
import { getThinkingLevels } from "../providers/thinkingLevels.js";
|
||||
import { DEFAULT_RETRY_CONFIG, HTTP_STATUS, resolveRetryEntry } from "../config/runtimeConfig.js";
|
||||
import { dbg } from "../utils/debugLog.js";
|
||||
@@ -25,6 +25,10 @@ const CODEX_SSE_USER_OUTPUT_PATTERNS = [
|
||||
];
|
||||
const CODEX_SSE_PEEK_BYTES = 256 * 1024;
|
||||
const CODEX_MODEL_CAPACITY_MESSAGE = "Selected model is at capacity. Please try a different model.";
|
||||
function isCodexResponsesLiteModel(model) {
|
||||
const baseId = String(model || "").replace(/\([^()]+\)\s*$/, "");
|
||||
return getProviderModels("cx").some((entry) => entry.id === baseId && entry.responsesLite === true);
|
||||
}
|
||||
|
||||
// Server-generated item id prefixes that Codex /responses cannot resolve when store=false
|
||||
const SERVER_ID_PATTERN = /^(rs|fc|resp|msg)_/;
|
||||
@@ -43,7 +47,7 @@ const CODEX_PASSTHROUGH_TOOL_TYPES = new Set(["custom"]);
|
||||
const RESPONSES_API_ALLOWLIST = new Set([
|
||||
"model", "input", "instructions", "tools", "tool_choice", "stream", "store",
|
||||
"reasoning", "service_tier", "include", "prompt_cache_key", "client_metadata",
|
||||
"text"
|
||||
"text", "parallel_tool_calls"
|
||||
]);
|
||||
|
||||
// Convert role=system → role=developer in body.input (keeps content in cacheable prefix)
|
||||
@@ -57,13 +61,14 @@ function convertSystemToDeveloperRole(body) {
|
||||
}
|
||||
|
||||
// Strip server-generated item IDs (rs_/fc_/resp_/msg_) from input — avoids 404 with store=false
|
||||
function stripStoredItemReferences(body) {
|
||||
function stripStoredItemReferences(body, preserveLitePrefix = false) {
|
||||
if (!Array.isArray(body.input)) return;
|
||||
body.input = body.input.filter((item) => {
|
||||
if (typeof item === "string" && SERVER_ID_PATTERN.test(item)) return false;
|
||||
if (item && typeof item === "object" && !Array.isArray(item)) {
|
||||
if (item.type === "item_reference") return false;
|
||||
if (typeof item.id === "string" && SERVER_ID_PATTERN.test(item.id)) delete item.id;
|
||||
if (typeof item.id === "string" && SERVER_ID_PATTERN.test(item.id)
|
||||
&& !(preserveLitePrefix && item.role === "developer" && item.id.startsWith("msg_"))) delete item.id;
|
||||
}
|
||||
return true;
|
||||
});
|
||||
@@ -138,6 +143,7 @@ function resolveCacheSessionId(body, credentials) {
|
||||
function normalizeReasoningEffort(model, value) {
|
||||
const supportedLevels = getThinkingLevels("codex", model);
|
||||
if (supportedLevels?.includes(value)) return value;
|
||||
if (isCodexResponsesLiteModel(model) && (value === "none" || value === "minimal")) return "low";
|
||||
if (value === "ultra" && supportedLevels?.includes("max")) return "max";
|
||||
if (value === "max" || value === "ultra") return "xhigh";
|
||||
return value;
|
||||
@@ -209,8 +215,11 @@ export class CodexExecutor extends BaseExecutor {
|
||||
* Override headers to add codex-specific identity headers.
|
||||
* transformRequest runs BEFORE buildHeaders, sets this._currentSessionId.
|
||||
*/
|
||||
buildHeaders(credentials, stream = true) {
|
||||
buildHeaders(credentials, stream = true, _url = null, model = null) {
|
||||
const headers = super.buildHeaders(credentials, stream);
|
||||
if (isCodexResponsesLiteModel(model && getModelUpstreamId("cx", model))) {
|
||||
headers["x-openai-internal-codex-responses-lite"] = "true";
|
||||
}
|
||||
headers["session_id"] = this._currentSessionId || credentials?.connectionId || "default";
|
||||
// Identify client type to Codex backend (matches official codex CLI)
|
||||
if (!headers["originator"]) headers["originator"] = "codex_cli_rs";
|
||||
@@ -408,6 +417,8 @@ export class CodexExecutor extends BaseExecutor {
|
||||
// Convert string input to array format (Codex API requires input as array)
|
||||
const normalized = normalizeResponsesInput(body.input);
|
||||
if (normalized) body.input = normalized;
|
||||
const upstreamModel = getModelUpstreamId("cx", body.model || model);
|
||||
const responsesLite = isCodexResponsesLiteModel(upstreamModel);
|
||||
|
||||
// Ensure input is present and non-empty (Codex API rejects empty input)
|
||||
if (!body.input || (Array.isArray(body.input) && body.input.length === 0)) {
|
||||
@@ -417,7 +428,7 @@ export class CodexExecutor extends BaseExecutor {
|
||||
// Keep system prompts in body.input as role=developer so they stay in the cacheable prefix
|
||||
convertSystemToDeveloperRole(body);
|
||||
// Strip server-generated item IDs (rs_/fc_/resp_/msg_) — Codex /responses can't resolve when store=false
|
||||
stripStoredItemReferences(body);
|
||||
stripStoredItemReferences(body, responsesLite);
|
||||
// Flatten function tools + drop unsupported types
|
||||
normalizeCodexTools(body);
|
||||
|
||||
@@ -425,7 +436,7 @@ export class CodexExecutor extends BaseExecutor {
|
||||
body.stream = true;
|
||||
|
||||
// If no instructions provided, inject default Codex instructions
|
||||
if (!body.instructions || body.instructions.trim() === "") {
|
||||
if (!responsesLite && (!body.instructions || body.instructions.trim() === "")) {
|
||||
body.instructions = CODEX_DEFAULT_INSTRUCTIONS;
|
||||
}
|
||||
|
||||
@@ -438,7 +449,29 @@ export class CodexExecutor extends BaseExecutor {
|
||||
}
|
||||
|
||||
// Map virtual Codex review models to the upstream Codex model before suffix parsing.
|
||||
body.model = getModelUpstreamId("cx", body.model || model);
|
||||
body.model = upstreamModel;
|
||||
|
||||
if (responsesLite) {
|
||||
// Codex 0.155 carries tools and instructions as input prefix items.
|
||||
const input = Array.isArray(body.input) ? body.input : [body.input];
|
||||
const hasLitePrefix = input.some((item) => item?.type === "additional_tools");
|
||||
if (!hasLitePrefix) {
|
||||
const instructions = typeof body.instructions === "string" && body.instructions.trim()
|
||||
? body.instructions : CODEX_DEFAULT_INSTRUCTIONS;
|
||||
const prefix = [{ type: "additional_tools", role: "developer", tools: Array.isArray(body.tools) ? body.tools : [] }];
|
||||
if (instructions) {
|
||||
prefix.push({ type: "message", role: "developer", content: [{ type: "input_text", text: instructions }] });
|
||||
}
|
||||
input.unshift(...prefix);
|
||||
}
|
||||
body.input = input;
|
||||
body.instructions = "";
|
||||
body.tools = null;
|
||||
body.tool_choice ||= "auto";
|
||||
body.parallel_tool_calls = false;
|
||||
} else {
|
||||
delete body.parallel_tool_calls;
|
||||
}
|
||||
|
||||
// Extract thinking level from model name suffix
|
||||
// e.g., gpt-5.3-codex-high → high, gpt-5.3-codex → medium (default)
|
||||
@@ -455,12 +488,13 @@ export class CodexExecutor extends BaseExecutor {
|
||||
|
||||
// Priority: explicit reasoning.effort > reasoning_effort param > model suffix > default (medium)
|
||||
if (!body.reasoning) {
|
||||
const effort = normalizeReasoningEffort(body.model, body.reasoning_effort || modelEffort || 'low');
|
||||
body.reasoning = { effort, summary: "auto" };
|
||||
const effort = normalizeReasoningEffort(body.model, body.reasoning_effort || modelEffort || (responsesLite ? 'medium' : 'low'));
|
||||
body.reasoning = responsesLite ? { effort } : { effort, summary: "auto" };
|
||||
} else {
|
||||
body.reasoning.effort = normalizeReasoningEffort(body.model, body.reasoning.effort);
|
||||
if (!body.reasoning.summary) body.reasoning.summary = "auto";
|
||||
if (!responsesLite && !body.reasoning.summary) body.reasoning.summary = "auto";
|
||||
}
|
||||
if (responsesLite) body.reasoning.context = "all_turns";
|
||||
delete body.reasoning_effort;
|
||||
|
||||
// Include reasoning encrypted content (required by Codex backend for reasoning models)
|
||||
|
||||
@@ -148,7 +148,7 @@ export async function inspectAndWrapCommandCodeResponse(originalResponse, model)
|
||||
const reader = originalResponse.body.getReader();
|
||||
const decoder = new TextDecoder();
|
||||
let buffer = "";
|
||||
const bufferedLines = [];
|
||||
const rawChunks = [];
|
||||
let detectedError = null;
|
||||
|
||||
try {
|
||||
@@ -162,16 +162,15 @@ export async function inspectAndWrapCommandCodeResponse(originalResponse, model)
|
||||
const parsed = JSON.parse(jsonStr);
|
||||
if (parsed?.type === "error") {
|
||||
detectedError = parsed;
|
||||
} else {
|
||||
bufferedLines.push(trimmed);
|
||||
}
|
||||
} catch {
|
||||
bufferedLines.push(trimmed);
|
||||
/* ignore */
|
||||
}
|
||||
}
|
||||
break;
|
||||
}
|
||||
|
||||
rawChunks.push(value);
|
||||
buffer += decoder.decode(value, { stream: true });
|
||||
const lines = buffer.split("\n");
|
||||
buffer = lines.pop() || "";
|
||||
@@ -182,7 +181,6 @@ export async function inspectAndWrapCommandCodeResponse(originalResponse, model)
|
||||
if (!trimmed) continue;
|
||||
const jsonStr = trimmed.startsWith("data:") ? trimmed.slice(5).trim() : trimmed;
|
||||
if (!jsonStr || jsonStr === "[DONE]") {
|
||||
bufferedLines.push(trimmed);
|
||||
stopLoop = true;
|
||||
break;
|
||||
}
|
||||
@@ -191,7 +189,6 @@ export async function inspectAndWrapCommandCodeResponse(originalResponse, model)
|
||||
try {
|
||||
event = JSON.parse(jsonStr);
|
||||
} catch {
|
||||
bufferedLines.push(trimmed);
|
||||
continue;
|
||||
}
|
||||
|
||||
@@ -201,8 +198,6 @@ export async function inspectAndWrapCommandCodeResponse(originalResponse, model)
|
||||
break;
|
||||
}
|
||||
|
||||
bufferedLines.push(trimmed);
|
||||
|
||||
if (
|
||||
event?.type === "text-delta" ||
|
||||
event?.type === "reasoning-delta" ||
|
||||
@@ -245,29 +240,18 @@ export async function inspectAndWrapCommandCodeResponse(originalResponse, model)
|
||||
);
|
||||
}
|
||||
|
||||
const combinedStream = createReplayedStream(bufferedLines, buffer, reader);
|
||||
const combinedStream = createRawReplayedStream(rawChunks, reader);
|
||||
return wrapNdjsonAsOpenAISse(combinedStream, model, originalResponse);
|
||||
}
|
||||
|
||||
function createReplayedStream(bufferedLines, remainingBuffer, reader) {
|
||||
const encoder = new TextEncoder();
|
||||
let replayed = false;
|
||||
function createRawReplayedStream(rawChunks, reader) {
|
||||
let chunkIndex = 0;
|
||||
|
||||
return new ReadableStream({
|
||||
async pull(controller) {
|
||||
if (!replayed) {
|
||||
replayed = true;
|
||||
let prefix = bufferedLines.join("\n");
|
||||
if (prefix && remainingBuffer) {
|
||||
prefix += "\n" + remainingBuffer;
|
||||
} else if (remainingBuffer) {
|
||||
prefix = remainingBuffer;
|
||||
} else if (prefix) {
|
||||
prefix += "\n";
|
||||
}
|
||||
if (prefix) {
|
||||
controller.enqueue(encoder.encode(prefix));
|
||||
}
|
||||
if (chunkIndex < rawChunks.length) {
|
||||
controller.enqueue(rawChunks[chunkIndex++]);
|
||||
return;
|
||||
}
|
||||
|
||||
try {
|
||||
|
||||
@@ -1,12 +1,13 @@
|
||||
import { BaseExecutor } from "./base.js";
|
||||
import { PROVIDERS, PROVIDER_OAUTH } from "../config/providers.js";
|
||||
import { ANTHROPIC_API_VERSION, OPENAI_COMPAT_BASE, ANTHROPIC_COMPAT_BASE, selectAnthropicBeta } from "../providers/shared.js";
|
||||
import { ANTHROPIC_API_VERSION, OPENAI_COMPAT_BASE, ANTHROPIC_COMPAT_BASE, selectAnthropicBeta, mergeAnthropicBeta } from "../providers/shared.js";
|
||||
import { resolveOpenAICompatibleApiType } from "../services/provider.js";
|
||||
import { OAUTH_ENDPOINTS, buildKimiHeaders } from "../config/appConstants.js";
|
||||
import { buildClineHeaders } from "../shared/clineAuth.js";
|
||||
import { proxyAwareFetch } from "../utils/proxyFetch.js";
|
||||
import { injectReasoningContent } from "../utils/reasoningContentInjector.js";
|
||||
import { stripUnsupportedParams } from "../translator/concerns/paramSupport.js";
|
||||
import { extractClaudeSessionIdFromUserId } from "../utils/claudeCloaking.js";
|
||||
|
||||
// Auth header descriptors — derived from registry transport.auth, fallback to hardcoded defaults.
|
||||
const BEARER = { combined: true, header: "Authorization", scheme: "bearer" };
|
||||
@@ -164,9 +165,21 @@ export class DefaultExecutor extends BaseExecutor {
|
||||
// a node fronting Kimi or GLM answers on its own ids and never matches, so
|
||||
// gateways that would choke on unknown beta flags are left untouched.
|
||||
const isClaudeModel = typeof model === "string" && /^claude-/.test(model);
|
||||
const clientBeta = credentials?.rawHeaders?.["anthropic-beta"];
|
||||
if (model && (this.provider === "claude"
|
||||
|| (this.provider?.startsWith?.("anthropic-compatible-") && isClaudeModel))) {
|
||||
headers["Anthropic-Beta"] = selectAnthropicBeta(model, body);
|
||||
headers["Anthropic-Beta"] = mergeAnthropicBeta(selectAnthropicBeta(model, body), clientBeta);
|
||||
} else if (this.provider === "anthropic" && clientBeta) {
|
||||
headers["Anthropic-Beta"] = mergeAnthropicBeta(headers["Anthropic-Beta"], clientBeta);
|
||||
}
|
||||
|
||||
// Claude OAuth: align x-claude-code-session-id with metadata.user_id.session_id if missing
|
||||
if (this.provider === "claude" && !headers["x-claude-code-session-id"]) {
|
||||
const token = credentials?.accessToken || credentials?.apiKey || "";
|
||||
if (token.includes("sk-ant-oat")) {
|
||||
const sid = extractClaudeSessionIdFromUserId(body?.metadata?.user_id);
|
||||
if (sid) headers["x-claude-code-session-id"] = sid;
|
||||
}
|
||||
}
|
||||
|
||||
// Strip first-party Claude Code identity headers for non-Anthropic anthropic-compatible upstreams
|
||||
|
||||
@@ -1,8 +1,8 @@
|
||||
import crypto from "node:crypto";
|
||||
import { DefaultExecutor } from "./default.js";
|
||||
import { resolveSessionId } from "../utils/sessionManager.js";
|
||||
import { modelTargetFormat } from "../providers/models/schema.js";
|
||||
import { getProviderModels } from "../config/providerModels.js";
|
||||
import { getModelTargetFormat } from "../config/providerModels.js";
|
||||
import { FORMATS } from "../translator/formats.js";
|
||||
import {
|
||||
normalizeResponsesInput,
|
||||
clampResponsesCallId,
|
||||
@@ -41,16 +41,10 @@ function translatedSession(sessionId, clientTool) {
|
||||
return `ses_${digest}`;
|
||||
}
|
||||
|
||||
// Strip the thinking suffix "model(level)" so checks hit the base id.
|
||||
function baseModelId(model) {
|
||||
return String(model || "").replace(/\([^()]+\)\s*$/, "").trim();
|
||||
}
|
||||
|
||||
// Responses-only per the provider registry (grok-4.6, gpt-5.6-luna, muse-spark, …).
|
||||
// Reading the registry keeps this in sync with config — never hardcode model ids here.
|
||||
// Responses-only per the provider registry (grok-4.6, gpt-5.6-luna, muse-spark, …),
|
||||
// including the family-regex fallback for passthrough ids — never hardcode model ids here.
|
||||
function isResponsesModel(model) {
|
||||
const entry = getProviderModels("opencode-go").find((m) => m.id === baseModelId(model));
|
||||
return modelTargetFormat(entry) === "openai-responses";
|
||||
return getModelTargetFormat("opencode-go", model) === FORMATS.OPENAI_RESPONSES;
|
||||
}
|
||||
|
||||
// Flatten Chat Completions tool declarations into the Responses flat shape and
|
||||
|
||||
Reference in New Issue
Block a user