[{"_1":2,"_3":-5,"_4":-5},"loaderData",{"_5":6,"_7":8,"_9":10},"actionData","errors","root",{"_201":202,"_203":204},"routes/_app.x.$handle.$name",{"_106":105,"_39":52,"_114":96,"_34":115,"_13":14,"_116":-5,"_11":117,"_118":119,"_120":121,"_122":123,"_124":125,"_26":-5,"_126":-5,"_127":-5,"_128":33,"_129":130,"_131":31,"_132":-5,"_133":-5,"_134":135},"routes/_app.x.$handle.$name.branch.$branch.version.$version.diff",{"_11":12,"_13":14,"_15":16,"_17":18,"_19":-5},"project",{"_50":51,"_39":52,"_45":53,"_54":-5,"_55":33,"_56":57,"_58":57,"_59":60,"_61":-5,"_19":62,"_63":64,"_65":31,"_66":-5,"_67":-5,"_68":69,"_70":71,"_72":73,"_74":75,"_76":77,"_78":-5,"_79":80,"_81":80,"_82":83,"_84":83,"_85":83,"_86":80,"_87":80,"_88":89,"_90":91,"_92":93,"_94":33},"version",34,"changes",[25],"aiChatInfo",{"_20":21,"_22":-5,"_23":-5,"_24":-5},"userId","aiChatMessageId","4f252adc-9e65-426f-9eb1-64f13ac56340","aiChatThreadId","aiChatThreadUserId","aiChatThreadTitle",{"_26":27,"_28":29,"_30":31,"_32":33,"_34":35,"_36":35},"file",{"_37":38,"_39":35,"_40":41,"_42":44,"_45":46,"_47":14,"_48":49,"_30":33},"previous",{"_37":38,"_39":35,"_40":41,"_42":43},"deleted",true,"deletedHere",false,"path","alice.ts","latestPath","fileId","e3cbb9be-841a-11f1-b603-1607ee4eb77e","name","fileType","script","content","// Alice WebSocket client — connects to alice.yandex.ru via uniproxy\n// and provides a simple async interface for sending text messages and\n// receiving streaming responses.\n\nconst UNIPROXY_URL = \"wss://uniproxy.alice.yandex.ru/uni.ws\";\nconst ALICE_PAGE_URL = \"https://alice.yandex.ru/\";\nconst APP_ID = \"ru.yandex.webstandalone.desktop\";\n\nexport interface AliceSession {\n sk: string;\n uuid: string;\n cookies: string;\n}\n\nexport interface AliceResponseChunk {\n text: string;\n done: boolean;\n raw: unknown;\n}\n\n/** Fetch alice.yandex.ru to get a fresh anonymous session (sk, uuid, cookies). */\nexport async function getSession(): Promise {\n const resp = await fetch(ALICE_PAGE_URL, {\n headers: {\n \"User-Agent\":\n \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120.0.0.0 Safari/537.36\",\n \"Accept\": \"text/html,application/xhtml+xml\",\n \"Accept-Language\": \"ru-RU,ru;q=0.9,en;q=0.8\",\n },\n });\n\n const html = await resp.text();\n const sk = html.match(/\"sk\":\"([^\"]+)\"/)?.[1] ?? \"\";\n const uuid = html.match(/\"uuid\":\"([^\"]+)\"/)?.[1] ?? \"\";\n const setCookies = resp.headers.getSetCookie?.() ?? [];\n const cookies = setCookies.map((c) => c.split(\";\")[0]).join(\"; \");\n\n if (!sk || !uuid) {\n throw new Error(\"Failed to extract session data from alice.yandex.ru\");\n }\n\n return { sk, uuid, cookies };\n}\n\nfunction genId(): string {\n return Array.from(crypto.getRandomValues(new Uint8Array(16)), (b) =>\n b.toString(16).padStart(2, \"0\")\n ).join(\"\");\n}\n\n/**\n * Send a text message to Alice and receive streaming response chunks.\n * Handles WebSocket connection, auth, and response parsing.\n */\nexport async function* sendToAlice(\n text: string,\n session: AliceSession,\n dialogId = \"\",\n): AsyncGenerator {\n const { sk, uuid } = session;\n let seqNum = 0;\n\n // Connect WebSocket\n const ws = new WebSocket(UNIPROXY_URL);\n\n // Create a queue for incoming messages\n const messageQueue: string[] = [];\n let wsClosed = false;\n let wsError: Error | null = null;\n\n ws.onmessage = (event) => {\n messageQueue.push(String(event.data));\n };\n\n ws.onerror = () => {\n wsError = new Error(\"WebSocket error\");\n };\n\n ws.onclose = () => {\n wsClosed = true;\n };\n\n // Wait for connection\n await new Promise((resolve, reject) => {\n const timeout = setTimeout(() => reject(new Error(\"WS connect timeout\")), 10000);\n\n ws.onopen = () => {\n clearTimeout(timeout);\n resolve();\n };\n\n ws.onerror = () => {\n clearTimeout(timeout);\n reject(new Error(\"WS connection failed\"));\n };\n });\n\n // Step 1: Send auth (SynchronizeState)\n ws.send(JSON.stringify({\n event: {\n header: {\n namespace: \"System\",\n name: \"SynchronizeState\",\n messageId: genId(),\n seqNumber: seqNum++,\n },\n payload: {\n accept_invalid_auth: true,\n auth_token: sk,\n uuid,\n vins: {\n application: {\n app_id: APP_ID,\n platform: \"desktop\",\n device_id: uuid,\n device_model: \"desktop\",\n },\n },\n supported_features: [\n \"supports_streaming_response\",\n \"supports_markdown_response\",\n \"supports_rich_json_cards\",\n \"supports_default_dialog_as_dedicated\",\n \"print_text_in_message_view\",\n ],\n request: { experiments: [] },\n speechkitVersion: \"4.16.7\",\n },\n },\n }));\n\n // Wait for auth response\n let authenticated = false;\n const authDeadline = Date.now() + 5000;\n while (!authenticated && Date.now() < authDeadline && !wsClosed) {\n if (messageQueue.length > 0) {\n const msg = messageQueue.shift()!;\n if (msg.includes(\"SynchronizeStateResponse\")) {\n authenticated = true;\n }\n } else {\n await new Promise((r) => setTimeout(r, 50));\n }\n }\n\n if (!authenticated) {\n try { ws.close(); } catch {}\n throw new Error(\"Authentication failed\");\n }\n\n // Step 2: Send TextInput\n const requestId = genId();\n const messageId = genId();\n const now = new Date();\n const clientTime = now.toISOString().slice(0, 19).replace(/[:+-]/g, \"\");\n const timestamp = String(Date.now()).slice(0, -3);\n\n ws.send(JSON.stringify({\n event: {\n header: {\n namespace: \"Vins\",\n name: \"TextInput\",\n messageId,\n seqNumber: seqNum++,\n },\n payload: {\n application: {\n app_id: APP_ID,\n app_version: \"1.114.1\",\n platform: \"desktop\",\n os_version: \"linux\",\n uuid,\n device_id: uuid,\n lang: \"ru-RU\",\n client_time: clientTime,\n timezone: \"Europe/Moscow\",\n timestamp,\n device_model: \"desktop\",\n },\n header: {\n prev_req_id: null,\n request_id: requestId,\n dialog_id: dialogId,\n dialog_type: \"DEDICATED_CHAT\",\n },\n request: {\n event: {\n type: \"text_input\",\n text: text,\n },\n voice_session: false,\n experiments: [],\n uniproxy_options: {\n background_response_streaming_options: {\n request_with_shown_new_dialog: {},\n },\n },\n additional_options: {\n origin_domain: \"yandex.ru\",\n supported_features: [],\n unsupported_features: [],\n },\n reset_session: true,\n },\n },\n },\n }));\n\n // Step 3: Read response messages\n let fullText = \"\";\n let responseDone = false;\n const responseDeadline = Date.now() + 30000;\n\n while (!responseDone && Date.now() < responseDeadline && !wsClosed) {\n if (wsError) {\n try { ws.close(); } catch {}\n throw wsError;\n }\n\n if (messageQueue.length > 0) {\n const rawMsg = messageQueue.shift()!;\n let parsed: any;\n try {\n parsed = JSON.parse(rawMsg);\n } catch {\n continue;\n }\n\n // Skip ack messages\n if (rawMsg.includes(\"InputStartAck\")) continue;\n if (rawMsg.includes(\"SynchronizeStateResponse\")) continue;\n\n // Check for VinsResponse (final response)\n const directive = parsed.directive;\n if (!directive) continue;\n\n const headerName = directive.header?.name;\n const payload = directive.payload;\n\n // Error\n if (headerName === \"EventException\") {\n try { ws.close(); } catch {}\n throw new Error(\n payload?.error?.message ?? \"Alice API error\",\n );\n }\n\n // VinsResponse — contains the response text\n if (headerName === \"VinsResponse\") {\n const response = payload?.response;\n if (response) {\n // Extract text from the response\n // The response can have json_response with text, or output_speech with text\n const text = response.json_response?.text ??\n response.output_speech?.text ??\n response.text ??\n \"\";\n\n if (text && text !== fullText) {\n // Streaming: yield the delta\n const delta = text.slice(fullText.length);\n fullText = text;\n if (delta) {\n yield { text: delta, done: false, raw: parsed };\n }\n }\n\n // Check if streaming is done\n if (!response.is_streaming) {\n responseDone = true;\n // Final yield with remaining text if any\n if (fullText === \"\") {\n yield { text: text, done: true, raw: parsed };\n } else {\n yield { text: \"\", done: true, raw: parsed };\n }\n }\n }\n }\n\n // DeferredAliceResponse — also contains response data\n if (headerName === \"DeferredAliceResponse\") {\n const response = payload?.response;\n if (response?.text && response.text !== fullText) {\n const delta = response.text.slice(fullText.length);\n fullText = response.text;\n if (delta) {\n yield { text: delta, done: false, raw: parsed };\n }\n }\n }\n\n // DialogControl — might indicate end of response\n if (headerName === \"DialogControl\") {\n if (payload?.is_final || payload?.finished) {\n responseDone = true;\n yield { text: \"\", done: true, raw: parsed };\n }\n }\n } else {\n await new Promise((r) => setTimeout(r, 50));\n }\n }\n\n // If we haven't received a \"done\" signal but have text, yield it\n if (!responseDone && fullText) {\n yield { text: \"\", done: true, raw: null };\n }\n\n try { ws.close(); } catch {}\n}\n\n/** Send a text message to Alice and get the complete response (non-streaming). */\nexport async function askAlice(\n text: string,\n session?: AliceSession,\n): Promise {\n const sess = session ?? await getSession();\n let fullResponse = \"\";\n for await (const chunk of sendToAlice(text, sess)) {\n fullResponse += chunk.text;\n }\n return fullResponse;\n}\n","// Alice WebSocket client — connects to alice.yandex.ru via uniproxy\n// and provides a simple async interface for sending text messages and\n// receiving streaming responses.\n\nconst UNIPROXY_URL = \"wss://uniproxy.alice.yandex.ru/uni.ws\";\nconst ALICE_PAGE_URL = \"https://alice.yandex.ru/\";\nconst APP_ID = \"ru.yandex.webstandalone.desktop\";\n\nexport interface AliceSession {\n sk: string;\n uuid: string;\n cookies: string;\n}\n\nexport interface AliceResponseChunk {\n text: string;\n done: boolean;\n raw: unknown;\n}\n\n/** Fetch alice.yandex.ru to get a fresh anonymous session (sk, uuid, cookies). */\nexport async function getSession(): Promise {\n const resp = await fetch(ALICE_PAGE_URL, {\n headers: {\n \"User-Agent\":\n \"Mozilla/5.0 (Windows NT 10.0; Win64; x64) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/120.0.0.0 Safari/537.36\",\n \"Accept\": \"text/html,application/xhtml+xml\",\n \"Accept-Language\": \"ru-RU,ru;q=0.9,en;q=0.8\",\n },\n });\n\n const html = await resp.text();\n const sk = html.match(/\"sk\":\"([^\"]+)\"/)?.[1] ?? \"\";\n const uuid = html.match(/\"uuid\":\"([^\"]+)\"/)?.[1] ?? \"\";\n const setCookies = resp.headers.getSetCookie?.() ?? [];\n const cookies = setCookies.map((c) => c.split(\";\")[0]).join(\"; \");\n\n if (!sk || !uuid) {\n throw new Error(\"Failed to extract session data from alice.yandex.ru\");\n }\n\n return { sk, uuid, cookies };\n}\n\nfunction genId(): string {\n return Array.from(crypto.getRandomValues(new Uint8Array(16)), (b) =>\n b.toString(16).padStart(2, \"0\")\n ).join(\"\");\n}\n\n/**\n * Send a text message to Alice and receive streaming response chunks.\n * Handles WebSocket connection, auth, and response parsing.\n */\nexport async function* sendToAlice(\n text: string,\n session: AliceSession,\n dialogId = \"\",\n): AsyncGenerator {\n const { sk, uuid } = session;\n let seqNum = 0;\n\n // Connect WebSocket\n const ws = new WebSocket(UNIPROXY_URL);\n\n // Create a queue for incoming messages\n const messageQueue: string[] = [];\n let wsClosed = false;\n let wsError: Error | null = null;\n\n ws.onmessage = (event) => {\n messageQueue.push(String(event.data));\n };\n\n ws.onerror = () => {\n wsError = new Error(\"WebSocket error\");\n };\n\n ws.onclose = () => {\n wsClosed = true;\n };\n\n // Wait for connection\n await new Promise((resolve, reject) => {\n const timeout = setTimeout(() => reject(new Error(\"WS connect timeout\")), 10000);\n\n ws.onopen = () => {\n clearTimeout(timeout);\n resolve();\n };\n\n ws.onerror = () => {\n clearTimeout(timeout);\n reject(new Error(\"WS connection failed\"));\n };\n });\n\n // Step 1: Send auth (SynchronizeState)\n ws.send(JSON.stringify({\n event: {\n header: {\n namespace: \"System\",\n name: \"SynchronizeState\",\n messageId: genId(),\n seqNumber: seqNum++,\n },\n payload: {\n accept_invalid_auth: true,\n auth_token: sk,\n uuid,\n vins: {\n application: {\n app_id: APP_ID,\n platform: \"desktop\",\n device_id: uuid,\n device_model: \"desktop\",\n },\n },\n supported_features: [\n \"supports_streaming_response\",\n \"supports_markdown_response\",\n \"supports_rich_json_cards\",\n \"supports_default_dialog_as_dedicated\",\n \"print_text_in_message_view\",\n ],\n request: { experiments: [] },\n speechkitVersion: \"4.16.7\",\n },\n },\n }));\n\n // Wait for auth response\n let authenticated = false;\n const authDeadline = Date.now() + 5000;\n while (!authenticated && Date.now() < authDeadline && !wsClosed) {\n if (messageQueue.length > 0) {\n const msg = messageQueue.shift()!;\n if (msg.includes(\"SynchronizeStateResponse\")) {\n authenticated = true;\n }\n } else {\n await new Promise((r) => setTimeout(r, 50));\n }\n }\n\n if (!authenticated) {\n try { ws.close(); } catch {}\n throw new Error(\"Authentication failed\");\n }\n\n // Step 2: Send TextInput\n const requestId = genId();\n const messageId = genId();\n const now = new Date();\n const clientTime = now.toISOString().slice(0, 19).replace(/[:+-]/g, \"\");\n const timestamp = String(Date.now()).slice(0, -3);\n\n ws.send(JSON.stringify({\n event: {\n header: {\n namespace: \"Vins\",\n name: \"TextInput\",\n messageId,\n seqNumber: seqNum++,\n },\n payload: {\n application: {\n app_id: APP_ID,\n app_version: \"1.114.1\",\n platform: \"desktop\",\n os_version: \"linux\",\n uuid,\n device_id: uuid,\n lang: \"ru-RU\",\n client_time: clientTime,\n timezone: \"Europe/Moscow\",\n timestamp,\n device_model: \"desktop\",\n },\n header: {\n prev_req_id: null,\n request_id: requestId,\n dialog_id: dialogId,\n dialog_type: \"DEDICATED_CHAT\",\n },\n request: {\n event: {\n type: \"text_input\",\n text: text,\n },\n voice_session: false,\n experiments: [],\n uniproxy_options: {\n background_response_streaming_options: {\n request_with_shown_new_dialog: {},\n },\n },\n additional_options: {\n origin_domain: \"yandex.ru\",\n supported_features: [],\n unsupported_features: [],\n },\n reset_session: true,\n },\n },\n },\n }));\n\n // Step 3: Read response messages\n let fullText = \"\";\n let responseDone = false;\n const responseDeadline = Date.now() + 30000;\n\n while (!responseDone && Date.now() < responseDeadline && !wsClosed) {\n if (wsError) {\n try { ws.close(); } catch {}\n throw wsError;\n }\n\n if (messageQueue.length > 0) {\n const rawMsg = messageQueue.shift()!;\n let parsed: any;\n try {\n parsed = JSON.parse(rawMsg);\n } catch {\n continue;\n }\n\n // Skip ack messages\n if (rawMsg.includes(\"InputStartAck\")) continue;\n if (rawMsg.includes(\"SynchronizeStateResponse\")) continue;\n\n // Check for VinsResponse (final response)\n const directive = parsed.directive;\n if (!directive) continue;\n\n const headerName = directive.header?.name;\n const payload = directive.payload;\n\n // Error\n if (headerName === \"EventException\") {\n try { ws.close(); } catch {}\n throw new Error(\n payload?.error?.message ?? \"Alice API error\",\n );\n }\n\n // VinsResponse — contains the response text\n if (headerName === \"VinsResponse\") {\n const response = payload?.response;\n if (response) {\n // Extract text from the response — Alice puts text in card.text or cards[0].text\n const text = response.card?.text ??\n response.cards?.[0]?.text ??\n response.text ??\n response.output_speech?.text ??\n response.json_response?.text ??\n \"\";\n\n if (text && text !== fullText) {\n const delta = text.slice(fullText.length);\n fullText = text;\n if (delta) {\n yield { text: delta, done: false, raw: parsed };\n }\n }\n\n // Check if streaming is done (non-streaming = final response)\n if (!response.is_streaming) {\n responseDone = true;\n yield { text: \"\", done: true, raw: parsed };\n }\n }\n }\n\n // DeferredAliceResponse — also contains response data\n if (headerName === \"DeferredAliceResponse\") {\n const response = payload?.response;\n const text = response?.card?.text ??\n response?.cards?.[0]?.text ??\n response?.text ?? \"\";\n if (text && text !== fullText) {\n const delta = text.slice(fullText.length);\n fullText = text;\n if (delta) {\n yield { text: delta, done: false, raw: parsed };\n }\n }\n }\n\n // DialogControl — might indicate end of response\n if (headerName === \"DialogControl\") {\n if (payload?.is_final || payload?.finished) {\n responseDone = true;\n yield { text: \"\", done: true, raw: parsed };\n }\n }\n } else {\n await new Promise((r) => setTimeout(r, 50));\n }\n }\n\n // If we haven't received a \"done\" signal but have text, yield it\n if (!responseDone && fullText) {\n yield { text: \"\", done: true, raw: null };\n }\n\n try { ws.close(); } catch {}\n}\n\n/** Send a text message to Alice and get the complete response (non-streaming). */\nexport async function askAlice(\n text: string,\n session?: AliceSession,\n): Promise {\n const sess = session ?? await getSession();\n let fullResponse = \"\";\n for await (const chunk of sendToAlice(text, sess)) {\n fullResponse += chunk.text;\n }\n return fullResponse;\n}\n","createdAt",["D",1784538766809],"branchVersion","branchId","c0d1bd45-cfa4-4439-a58d-8b2dd40d5344","id","b14b03ef-0b5a-41c2-8f35-784598396b5f","openai-yandex-adapter",["D",1784537552501],"deletedAt","legacyEnvVars","privacy","public","httpPrivacy","description","OpenAI-compatible API adapter for Yandex GPT (free tier)","imageUrl","27af7c01-80a6-42d0-8a42-18f99ce8bb03","pinnedFileId","a536781a-841f-11f1-8ecd-1607ee4eb77e","pinPreview","pinMigrated","screenshotVersion","likeCount",0,"orgId","019f7ba5-7ad8-732a-82a4-e6072669dcc2","tursoName","b14b03ef0b5a41c28f35784598396b5f","tursoHostname","b14b03ef0b5a41c28f35784598396b5f-val-town.aws-us-east-2.turso.io","tursoMeasuredAt",["D",1789484509295],"pinned","userPerm","r","valPerm","blobPerm","rw","emailPerm","sqlitePerm","projectPerm","telemetryPerm","user",{"_106":105,"_113":109},"org",{"_106":105,"_107":105,"_108":109,"_19":62,"_110":111,"_112":33},"branch",{"_13":95,"_39":96,"_50":49,"_97":51,"_98":-5,"_99":-5,"_100":-5,"_101":-5,"_102":-5,"_103":52,"_104":105},"isOwned",136,"main","projectId","forkedBranchId","forkedBranchVersion","forkedProjectPrivacy","forkedProjectId","forkedProjectOrgId","projectName","userHandle","dmkrj","handle","displayName","avatar","https://lh3.googleusercontent.com/a/ACg8ocKVGzvzjbF7__cpkyYO7FVbjNDQ3JVNm0gId3KrMdC7DKByr4M=s96-c","tier","free","canRestrictVals","profileImageUrl","branchName","","adoptOrgId",{"_50":51,"_39":52,"_45":196,"_54":-5,"_55":33,"_56":57,"_58":57,"_59":60,"_61":-5,"_19":62,"_63":64,"_65":31,"_66":-5,"_67":-5,"_68":69,"_70":71,"_72":73,"_74":75,"_76":197,"_78":-5,"_79":80,"_81":80,"_82":83,"_84":83,"_85":83,"_86":80,"_87":80,"_88":198,"_90":199,"_92":200,"_94":33},"tree",[179,180,181,182],"files",[179,180,181,182],"counts",{"_169":69,"_170":156,"_171":69,"_172":173,"_174":175},"contents",[136,137,138,139],"upstreamProject","upstreamBranch","emptyEnvVars","flags",[],"hasHttpFiles","grants","orgs","customDomains",{},{"_50":162,"_39":163,"_37":164,"_143":-5,"_48":49,"_47":165,"_145":33,"_45":166,"_13":156,"_40":26,"_42":167,"_30":33,"_150":-5,"_34":168},{"_50":159,"_39":35,"_37":38,"_143":-5,"_48":49,"_47":14,"_145":33,"_45":160,"_13":156,"_40":41,"_42":44,"_30":33,"_150":-5,"_34":161},{"_50":152,"_39":153,"_37":154,"_143":-5,"_48":49,"_47":147,"_145":33,"_45":155,"_13":156,"_40":41,"_42":157,"_30":33,"_150":-5,"_34":158},{"_50":140,"_39":141,"_37":142,"_143":-5,"_48":49,"_47":144,"_145":33,"_45":146,"_13":147,"_40":148,"_42":149,"_30":33,"_150":-5,"_34":151},"2f6847f2-3b36-489e-a3fa-7e6af36f123a","main.ts","58926db8-8418-11f1-aac1-1607ee4eb77e","parentId",30,"latest",["D",1784538654785],5,"http","import { Hono } from \"npm:hono\";\nimport { askAlice, getSession, sendToAlice } from \"./alice.ts\";\nimport { buildChunk, toOpenAIError, toOpenAIModels, toOpenAIResponse } from \"./convert.ts\";\nimport type { OpenAIChatRequest } from \"./convert.ts\";\n\nconst app = new Hono();\n\napp.onError((err) => Promise.reject(err));\n\n// ── Routes ────────────────────────────────────────────────────\n\n// Health check\napp.get(\"/\", (c) => c.json({\n status: \"ok\",\n service: \"openai-yandex-adapter\",\n backend: \"alice.yandex.ru\",\n free: true,\n streaming: true,\n}));\n\n// List models (OpenAI-compatible)\napp.get(\"/v1/models\", (c) => c.json(toOpenAIModels()));\n\n// Chat completions — proxies to alice.yandex.ru via WebSocket\napp.post(\"/v1/chat/completions\", async (c) => {\n let body: OpenAIChatRequest;\n try {\n body = await c.req.json();\n } catch {\n return c.json(toOpenAIError(\"Invalid JSON in request body.\"), 400);\n }\n\n if (!body.messages?.length) {\n return c.json(\n toOpenAIError(\"messages is required and must not be empty.\", \"invalid_request_error\", \"messages\"),\n 400,\n );\n }\n\n // Combine all messages into a single text prompt for Alice\n // (Alice doesn't support multi-turn via simple text, so we concatenate)\n const prompt = body.messages\n .map((m) => {\n if (m.role === \"system\") return `[Системный промпт: ${m.content}]`;\n if (m.role === \"assistant\") return `[Ассистент: ${m.content}]`;\n return m.content;\n })\n .join(\"\\n\\n\");\n\n const model = body.model || \"alice\";\n const isStream = body.stream ?? false;\n\n // Get Alice session\n let session;\n try {\n session = await getSession();\n } catch (err) {\n return c.json(\n toOpenAIError(\n `Failed to get Alice session: ${err instanceof Error ? err.message : String(err)}`,\n \"api_error\",\n ),\n 502,\n );\n }\n\n // ── Non-streaming response ────────────────────────────────\n if (!isStream) {\n try {\n const responseText = await askAlice(prompt, session);\n return c.json(toOpenAIResponse(\n {\n alternatives: [{\n message: { role: \"assistant\", text: responseText },\n status: \"ALTERNATIVE_STATUS_FINAL\",\n }],\n usage: { inputTokens: \"0\", completionTokens: \"0\", totalTokens: \"0\" },\n },\n model,\n ));\n } catch (err) {\n return c.json(\n toOpenAIError(\n `Alice error: ${err instanceof Error ? err.message : String(err)}`,\n \"api_error\",\n ),\n 502,\n );\n }\n }\n\n // ── Streaming response (SSE) ──────────────────────────────\n const id = `chatcmpl-${crypto.randomUUID()}`;\n const created = Math.floor(Date.now() / 1000);\n const encoder = new TextEncoder();\n\n const stream = new ReadableStream({\n async start(controller) {\n try {\n // Send initial role chunk\n controller.enqueue(\n encoder.encode(\n buildChunk(id, created, model, { role: \"assistant\" }, null),\n ),\n );\n\n // Stream Alice's response\n for await (const chunk of sendToAlice(prompt, session)) {\n if (chunk.text) {\n controller.enqueue(\n encoder.encode(\n buildChunk(id, created, model, { content: chunk.text }, null),\n ),\n );\n }\n if (chunk.done) {\n controller.enqueue(\n encoder.encode(\n buildChunk(id, created, model, {}, \"stop\"),\n ),\n );\n }\n }\n\n // If no done signal was sent, send one\n controller.enqueue(\n encoder.encode(\n buildChunk(id, created, model, {}, \"stop\"),\n ),\n );\n\n controller.enqueue(encoder.encode(\"data: [DONE]\\n\\n\"));\n controller.close();\n } catch (err) {\n // Send error as a chunk, then close\n const errMsg = err instanceof Error ? err.message : String(err);\n controller.enqueue(\n encoder.encode(\n buildChunk(id, created, model, { content: `[Error: ${errMsg}]` }, \"stop\"),\n ),\n );\n controller.enqueue(encoder.encode(\"data: [DONE]\\n\\n\"));\n controller.close();\n }\n },\n });\n\n return new Response(stream, {\n headers: {\n \"Content-Type\": \"text/event-stream\",\n \"Cache-Control\": \"no-cache\",\n \"Connection\": \"keep-alive\",\n },\n });\n});\n\nexport default app.fetch;\n","label",[141],"35b7461b-16ad-44fc-bd67-dc09cc3bd8d5","convert.ts","7ffb52c0-8418-11f1-b5d0-1607ee4eb77e",["D",1784537712195],1,"// Type definitions and conversion functions between OpenAI and Yandex GPT formats\n\n// ── OpenAI types ──────────────────────────────────────────────\n\nexport interface OpenAIMessage {\n role: \"system\" | \"user\" | \"assistant\";\n content: string;\n}\n\nexport interface OpenAIChatRequest {\n model: string;\n messages: OpenAIMessage[];\n temperature?: number;\n max_tokens?: number;\n stream?: boolean;\n top_p?: number;\n frequency_penalty?: number;\n presence_penalty?: number;\n n?: number;\n stop?: string | string[];\n user?: string;\n}\n\nexport interface OpenAIChatResponse {\n id: string;\n object: \"chat.completion\";\n created: number;\n model: string;\n choices: Array<{\n index: number;\n message: { role: string; content: string };\n finish_reason: string;\n }>;\n usage: {\n prompt_tokens: number;\n completion_tokens: number;\n total_tokens: number;\n };\n}\n\nexport interface OpenAIError {\n error: {\n message: string;\n type: string;\n param: string | null;\n code: string | null;\n };\n}\n\nexport interface OpenAIModelsResponse {\n object: \"list\";\n data: Array<{\n id: string;\n object: \"model\";\n created: number;\n owned_by: string;\n }>;\n}\n\n// ── Yandex GPT types ──────────────────────────────────────────\n\ninterface YandexMessage {\n role: \"system\" | \"user\" | \"assistant\";\n text: string;\n}\n\ninterface YandexCompletionOptions {\n stream: boolean;\n temperature?: number;\n maxTokens?: string;\n topP?: number;\n}\n\nexport interface YandexChatRequest {\n modelUri: string;\n completionOptions: YandexCompletionOptions;\n messages: YandexMessage[];\n}\n\ninterface YandexAlternative {\n message: { role: string; text: string };\n status: string;\n}\n\ninterface YandexResponse {\n alternatives: YandexAlternative[];\n usage?: {\n inputTokens?: string;\n completionTokens?: string;\n totalTokens?: string;\n };\n modelVersion?: string;\n}\n\n// ── Constants ─────────────────────────────────────────────────\n\nexport const YANDEX_ENDPOINT =\n \"https://llm.api.cloud.yandex.net/llm/v1/chat/completions\";\n\nexport const YANDEX_MODELS = [\n { id: \"yandexgpt\", label: \"YandexGPT\" },\n { id: \"yandexgpt-lite\", label: \"YandexGPT Lite\" },\n];\n\n/** Default model when client doesn't specify one. */\nexport const DEFAULT_MODEL = \"yandexgpt-lite\";\n\n/**\n * Map OpenAI model names to Yandex model names.\n * Lets you point any OpenAI SDK directly at this adapter without changing model names.\n */\nconst MODEL_ALIASES: Record = {\n // OpenAI → Yandex\n \"gpt-4\": \"yandexgpt\",\n \"gpt-4o\": \"yandexgpt\",\n \"gpt-4-turbo\": \"yandexgpt\",\n \"gpt-3.5-turbo\": \"yandexgpt-lite\",\n \"gpt-3.5\": \"yandexgpt-lite\",\n // Short aliases\n \"lite\": \"yandexgpt-lite\",\n \"pro\": \"yandexgpt\",\n};\n\n/** Resolve a model name through the alias table. */\nexport function resolveModel(model: string): string {\n return MODEL_ALIASES[model] ?? model;\n}\n\n// ── Conversion functions ──────────────────────────────────────\n\n/** Build a Yandex modelUri from an OpenAI-style model name. */\nexport function buildModelUri(model: string, folderId: string): string {\n // Allow full URIs to pass through directly\n if (model.startsWith(\"gpt://\") || model.startsWith(\"cls://\")) {\n return model;\n }\n return `gpt://${folderId}/${model}/latest`;\n}\n\n/** Convert an OpenAI chat request to Yandex GPT format. */\nexport function toYandexRequest(\n req: OpenAIChatRequest,\n folderId: string,\n): YandexChatRequest {\n const completionOptions: YandexCompletionOptions = {\n stream: req.stream ?? false,\n };\n\n if (req.temperature !== undefined) {\n completionOptions.temperature = req.temperature;\n }\n if (req.max_tokens !== undefined) {\n completionOptions.maxTokens = String(req.max_tokens);\n }\n if (req.top_p !== undefined) {\n completionOptions.topP = req.top_p;\n }\n\n return {\n modelUri: buildModelUri(req.model, folderId),\n completionOptions,\n messages: req.messages.map((m) => ({\n role: m.role,\n text: m.content,\n })),\n };\n}\n\n/** Map Yandex alternative status to OpenAI finish_reason. */\nfunction toFinishReason(status: string): string {\n switch (status) {\n case \"ALTERNATIVE_STATUS_TRUNCATED_FINAL\":\n return \"length\";\n case \"ALTERNATIVE_STATUS_CONTENT_FILTER\":\n return \"content_filter\";\n default:\n return \"stop\";\n }\n}\n\n/** Convert a Yandex GPT response to OpenAI chat completion format. */\nexport function toOpenAIResponse(\n yandexResp: YandexResponse,\n model: string,\n): OpenAIChatResponse {\n const alt = yandexResp.alternatives?.[0];\n\n return {\n id: `chatcmpl-${crypto.randomUUID()}`,\n object: \"chat.completion\",\n created: Math.floor(Date.now() / 1000),\n model,\n choices: [\n {\n index: 0,\n message: {\n role: alt?.message?.role ?? \"assistant\",\n content: alt?.message?.text ?? \"\",\n },\n finish_reason: alt ? toFinishReason(alt.status) : \"stop\",\n },\n ],\n usage: {\n prompt_tokens: parseInt(yandexResp.usage?.inputTokens ?? \"0\", 10),\n completion_tokens: parseInt(\n yandexResp.usage?.completionTokens ?? \"0\",\n 10,\n ),\n total_tokens: parseInt(yandexResp.usage?.totalTokens ?? \"0\", 10),\n },\n };\n}\n\n/** Build an OpenAI-format SSE chunk for streaming. */\nexport function buildChunk(\n id: string,\n created: number,\n model: string,\n delta: Record,\n finishReason: string | null,\n): string {\n const chunk = {\n id,\n object: \"chat.completion.chunk\" as const,\n created,\n model,\n choices: [\n {\n index: 0,\n delta,\n finish_reason: finishReason,\n },\n ],\n };\n return `data: ${JSON.stringify(chunk)}\\n\\n`;\n}\n\n/** Convert a Yandex error response body to OpenAI error format. */\nexport function toOpenAIError(\n message: string,\n type = \"invalid_request_error\",\n code: string | null = null,\n): OpenAIError {\n return {\n error: { message, type, param: null, code },\n };\n}\n\n/** List available models in OpenAI /v1/models format. */\nexport function toOpenAIModels(): OpenAIModelsResponse {\n const created = Math.floor(Date.now() / 1000);\n return {\n object: \"list\",\n data: YANDEX_MODELS.map((m) => ({\n id: m.id,\n object: \"model\" as const,\n created,\n owned_by: \"yandex\",\n })),\n };\n}\n",[153],"9c6e8959-8881-4196-a375-19874d8db8f9",["D",1784538766809],[35],"b6750cef-ca51-42f9-b035-9183f7431888","README.md","973ae392-8418-11f1-99d5-1607ee4eb77e",9,["D",1784537745735],"# OpenAI → Yandex GPT API Adapter\n\nDrop-in adapter that lets any OpenAI-compatible client (SDKs, libraries, tools) use **Yandex GPT** as the backend. Works with the Yandex Cloud free tier.\n\n## How it works\n\n```\nClient (OpenAI SDK) ──→ This adapter ──→ Yandex GPT API\n OpenAI format ↕ translate Yandex format\n```\n\nThe adapter translates:\n- **Requests**: OpenAI `POST /v1/chat/completions` → Yandex `llm/v1/chat/completions`\n- **Responses**: Yandex format → OpenAI format\n- **Streaming**: Yandex SSE (cumulative text) → OpenAI SSE (deltas)\n- **Model names**: OpenAI model names auto-mapped to Yandex models\n\n## Endpoints\n\n| Method | Path | Description |\n|--------|------|-------------|\n| `GET` | `/` | Health check |\n| `GET` | `/v1/models` | List available Yandex models |\n| `POST` | `/v1/chat/completions` | Chat completions (streaming + non-streaming) |\n\n## Setup\n\n### 1. Get Yandex Cloud credentials (free tier)\n\nYou need a **folder ID** and an **API key** from Yandex Cloud:\n\n1. Go to [Yandex Cloud Console](https://console.yandex.cloud/)\n2. Create or select a folder, copy its **Folder ID**\n3. Create a **Service Account** → generate an **API Key**\n4. No payment method required for the free tier\n\n### 2. Set environment variables\n\n- `YANDEX_FOLDER_ID` — your Yandex Cloud folder ID\n- `YANDEX_API_KEY` — your Yandex Cloud API key\n\n👉 Add YANDEX_FOLDER_ID here: https://www.val.town/x/dmkrj/openai-yandex-adapter/environment-variables?key=YANDEX_FOLDER_ID\n\n👉 Add YANDEX_API_KEY here: https://www.val.town/x/dmkrj/openai-yandex-adapter/environment-variables?key=YANDEX_API_KEY\n\nYou can also pass credentials per-request:\n- `Authorization: Bearer ` header\n- `x-folder-id: ` header\n\n### 3. Use with any OpenAI client\n\n```python\nfrom openai import OpenAI\n\nclient = OpenAI(\n base_url=\"https://dmkrj--58926db8841811f1aac11607ee4eb77e.web.val.run/v1\",\n api_key=\"\",\n)\n\nresponse = client.chat.completions.create(\n model=\"yandexgpt-lite\", # or \"gpt-3.5-turbo\" — auto-mapped!\n messages=[\n {\"role\": \"system\", \"content\": \"You are a helpful assistant.\"},\n {\"role\": \"user\", \"content\": \"Hello!\"},\n ],\n)\nprint(response.choices[0].message.content)\n```\n\n```javascript\nimport OpenAI from \"openai\";\n\nconst client = new OpenAI({\n baseURL: \"https://dmkrj--58926db8841811f1aac11607ee4eb77e.web.val.run/v1\",\n apiKey: \"\",\n});\n\nconst response = await client.chat.completions.create({\n model: \"yandexgpt-lite\",\n messages: [{ role: \"user\", content: \"Hello!\" }],\n});\nconsole.log(response.choices[0].message.content);\n```\n\n**Streaming** works too — just set `stream: true`:\n\n```python\nstream = client.chat.completions.create(\n model=\"yandexgpt-lite\",\n messages=[{\"role\": \"user\", \"content\": \"Write a haiku about coding.\"}],\n stream=True,\n)\nfor chunk in stream:\n print(chunk.choices[0].delta.content or \"\", end=\"\", flush=True)\n```\n\n## Model aliases\n\nThe adapter automatically maps OpenAI model names to Yandex models, so you can use existing OpenAI SDK code without changes:\n\n| OpenAI model name | Yandex model |\n|-------------------|-------------|\n| `gpt-4` | `yandexgpt` |\n| `gpt-4o` | `yandexgpt` |\n| `gpt-4-turbo` | `yandexgpt` |\n| `gpt-3.5-turbo` | `yandexgpt-lite` |\n| `gpt-3.5` | `yandexgpt-lite` |\n| `yandexgpt` | `yandexgpt` |\n| `yandexgpt-lite` | `yandexgpt-lite` |\n| *(omitted)* | `yandexgpt-lite` (default) |\n\nYou can also pass a full Yandex model URI directly (e.g. `gpt:///yandexgpt/rc1`).\n\n## Supported parameters\n\n| OpenAI param | Yandex equivalent | Notes |\n|--------------|-------------------|-------|\n| `model` | `modelUri` | Auto-mapped via aliases, defaults to `yandexgpt-lite` |\n| `messages` | `messages` | `content` → `text` |\n| `temperature` | `temperature` | Passed through |\n| `max_tokens` | `maxTokens` | Converted to string |\n| `top_p` | `topP` | Passed through |\n| `stream` | `stream` | SSE delta conversion |\n\n## Error handling\n\nThe adapter maps Yandex errors to OpenAI-compatible error format:\n\n| Yandex status | Adapter status | Error type |\n|---------------|---------------|------------|\n| 401 / 403 | 401 | `authentication_error` |\n| 429 | 429 | `rate_limit_error` |\n| 5xx | 502 | `api_error` |\n| 4xx | 400 | `api_error` |\n\n## Architecture\n\n```mermaid\ngraph LR\n A[OpenAI Client] -->|POST /v1/chat/completions| B[Adapter - Hono]\n B -->|resolveModel + toYandexRequest| C[Convert]\n C -->|fetch| D[Yandex GPT API]\n D -->|Response / SSE| B\n B -->|toOpenAIResponse| E[OpenAI-format JSON / SSE]\n E --> A\n```\n\n## Files\n\n- `main.ts` — HTTP handler (Hono routes, auth, streaming, rate-limit handling)\n- `convert.ts` — Type definitions, model aliases, and format conversion functions\n",[163],"pulls","branches","forks","envVars",{"_176":69,"_178":69},"connections",{"_176":69,"_177":69},"count","unfilledCount","emptyCount",{"_50":162,"_39":163,"_37":164,"_143":-5,"_48":49,"_47":165,"_145":33,"_45":193,"_13":156,"_40":26,"_42":167,"_30":33,"_150":-5,"_184":194,"_34":195},{"_50":159,"_39":35,"_37":38,"_143":-5,"_48":49,"_47":14,"_145":33,"_45":190,"_13":156,"_40":41,"_42":44,"_30":33,"_150":-5,"_184":191,"_34":192},{"_50":152,"_39":153,"_37":154,"_143":-5,"_48":49,"_47":147,"_145":33,"_45":187,"_13":156,"_40":41,"_42":157,"_30":33,"_150":-5,"_184":188,"_34":189},{"_50":140,"_39":141,"_37":142,"_143":-5,"_48":49,"_47":144,"_145":33,"_45":183,"_13":147,"_40":148,"_42":149,"_30":33,"_150":-5,"_184":185,"_34":186},["D",1784538654785],"children",[],[141],["D",1784537712195],[],[153],["D",1784538766809],[],[35],["D",1784537745735],[],[163],["D",1784537552501],["D",1789484509295],{"_106":105,"_113":109},{"_106":105,"_107":105,"_108":109,"_19":62,"_110":111,"_112":33},{"_13":95,"_39":96,"_50":49,"_97":51,"_98":-5,"_99":-5,"_100":-5,"_101":-5,"_102":-5,"_103":52,"_104":105},"ctx",{"_88":-5,"_133":234,"_235":236,"_237":238,"_239":33,"_240":33,"_241":33,"_242":31,"_243":-5,"_244":-7,"_245":-7,"_246":69,"_247":33,"_248":33,"_249":-5,"_250":33,"_251":33,"_252":33,"_253":-5},"ENV",{"_205":206,"_207":208,"_209":96,"_210":211,"_212":213,"_214":215,"_216":217,"_218":219,"_220":221,"_222":223,"_224":225,"_226":227,"_228":229,"_230":231,"_232":233},"REACT_APP_SENTRY_DSN","https://586020d86f05489db63a554690acf1e9@o1398359.ingest.sentry.io/6724499","RENDER_GIT_COMMIT","e87eb04e707cf64e7e532c8e8ec395f209e17f15","RENDER_GIT_BRANCH","POSTHOG_API_TOKEN","phc_nmoyuAkStEZuUp69rYlg1DZgrIt19Gx0Y4Hwf8tGgJP","POSTHOG_API_HOST","https://wart.val.town","WEB_API_TEMPLATE","https://{val}.web.val.run","APP_URL","https://www.val.town","MODULE_URL","https://esm.town","VANITY_WEB_API_TEMPLATE","https://{label}.val.run/","DENOLS_URL","https://lsp.val.town","NODE_ENV","production","CLOUDFLARE_TURNSTILE_SITE_KEY","0x4AAAAAACtaIZJH7s9rBMZ6","GOOGLE_CLIENT_ID","875989696899-jtsqkhlbdv5jrlsk8631p7ipjg3c6o62.apps.googleusercontent.com","STRIPE_PUBLISHABLE_KEY","pk_live_51N2zvWEpZTTLipIkVEwXrUG9AT0yeOOEYxszarZT1jpsZV58Ly6wlg3SltZpHN26CUalOBmUSY7adQRQiSkCZz2500qSXNkf8Y","API_URL","https://api.val.town",[],"preferencesKeybindings","default","preferencesTheme","light","preferencesFixers","preferencesCopilot","preferencesSelectionAi","preferencesCloseBrackets","preferencesChatPanelState","preferencesDprint","preferencesDenofmt","notificationCount","flagDenols","isBusinessOrgMember","businessOrgHandle","isBusinessOrgAdmin","isHigherTownieTier","navCollapsed","currentOrgId"]

Diff · v34