GitHub Models troubleshooting

[yayoimizuha]

What happened?

# librechat.yaml

endpoints:
  azureOpenAI:
    titleModel: "gpt-4.1"
    plugins: false
    assistants: false
    groups:
      - group: "Azure GitHub Models"
        apiKey: "${GITHUB_PAT}"
        baseURL: "https://models.github.ai/inference"
        serverless: true
        models:
          AI21_Jamba_1.5_Large:
            deploymentName: "ai21-labs/AI21-Jamba-1.5-Large"
          gpt-4.1:
            deploymentName: "openai/gpt-4.1"
  custom:
    - name: "GitHub Models"
      apiKey: "${GITHUB_PAT}"
      baseURL: "https://models.github.ai/inference"
      models:
        fetch: false
        default: [
          "openai/gpt-4.1",
          "ai21-labs/AI21-Jamba-1.5-Large"
        ]
      summarize: true

In both cases, the GPT-4.1 model worked correctly, but the AI21-Jamba-1.5-Large does not work.

# error for `custom` config
Something went wrong. Here's the specific error message we encountered: An error occurred while processing the request: 422 {"error":["Input should be 'tool'","Field required","Input should be 'system'","Input should be 'assistant'","Input should be a valid string"]}

# error for `azureOpenAI` config
Something went wrong. Here's the specific error message we encountered: An error occurred while processing the request: 404 Unknown model: /AI21_Jamba_1.5_Large Troubleshooting URL: https://js.langchain.com/docs/troubleshooting/errors/MODEL_NOT_FOUND/ 

In addition, all the curl commands indicated as samples are working.

curl -X POST "https://models.github.ai/inference/chat/completions" \
    -H "Content-Type: application/json" \
    -H "Authorization: Bearer $GITHUB_TOKEN" \
    -d '{
        "messages": [
            {
                "role": "system",
                "content": "You are a helpful assistant."
            },
            {
                "role": "user",
                "content": "What is the capital of France?"
            }
        ],
        "temperature": 1.0,
        "top_p": 1.0,
        "max_tokens": 1000,
        "model": "ai21-labs/AI21-Jamba-1.5-Large"
    }'

Version Information

LibreChat version:

$ docker logs -f LibreChat

> [email protected] backend
> cross-env NODE_ENV=production node api/server/index.js

Steps to Reproduce

Use GitHub Personal Access Token without any permissons and set into ${GITHUB_PAT}.

# librechat.yaml
interface:
  runCode: false
endpoints:
  azureOpenAI:
    titleModel: "gpt-4.1"
    plugins: false
    assistants: false
    groups:
      - group: "Azure GitHub Models"
        apiKey: "${GITHUB_PAT}"
        baseURL: "https://models.github.ai/inference"
        serverless: true
        models:
          AI21_Jamba_1.5_Large:
            deploymentName: "ai21-labs/AI21-Jamba-1.5-Large"
          gpt-4.1:
            deploymentName: "openai/gpt-4.1"
  custom:
    - name: "GitHub Models"
      apiKey: "${GITHUB_PAT}"
      baseURL: "https://models.github.ai/inference"
      models:
        fetch: false
        default: [
          "openai/gpt-4.1",
          "ai21-labs/AI21-Jamba-1.5-Large"
        ]
      titleModel: "microsoft/Phi-4-mini-instruct"
      summarize: true

What browsers are you seeing the problem on?

Firefox

Relevant log output

docker logs -f LibreChat

> [email protected] backend
> cross-env NODE_ENV=production node api/server/index.js

2025-07-26 14:45:45 info: [Optional] Redis not initialized.
2025-07-26 14:45:46 info: [Optional] IoRedis not initialized for rate limiters.
2025-07-26 14:45:46 info: Connected to MongoDB
2025-07-26 14:45:46 info: [indexSync] Starting index synchronization check...
2025-07-26 14:45:46 info: Custom config file loaded:
2025-07-26 14:45:46 info: {
  "version": "1.1.5",
  "cache": true,
  "speech": {
    "tts": {
      "openai": {
        "url": "",
        "apiKey": "${TTS_API_KEY}",
        "model": "tts-1-hd",
        "voices": [
          "alloy",
          "echo",
          "fable",
          "onyx",
          "nova",
          "shimmer"
        ]
      }
    },
    "stt": {
      "openai": {
        "url": "",
        "apiKey": "${STT_API_KEY}",
        "model": "whisper-1"
      }
    }
  },
  "interface": {
    "runCode": false
  },
  "endpoints": {
    "azureOpenAI": {
      "titleModel": "gpt-4.1",
      "plugins": false,
      "assistants": false,
      "groups": [
        {
          "group": "Azure GitHub Models",
          "apiKey": "${GITHUB_PAT}",
          "baseURL": "https://models.github.ai/inference",
          "serverless": true,
          "models": {
            "AI21_Jamba_1.5_Large": {
              "deploymentName": "ai21-labs/AI21-Jamba-1.5-Large"
            },
            "gpt-4.1": {
              "deploymentName": "openai/gpt-4.1"
            }
          }
        }
      ]
    },
    "custom": [
      {
        "name": "GitHub Models",
        "apiKey": "${GITHUB_PAT}",
        "baseURL": "https://models.github.ai/inference",
        "models": {
          "fetch": false,
          "default": [
            "openai/gpt-4.1",
            "ai21-labs/AI21-Jamba-1.5-Large"
          ]
        },
        "titleModel": "microsoft/Phi-4-mini-instruct",
        "summarize": true
      },
      {
        "name": "xAI",
        "apiKey": "${XAI_API_KEY}",
        "baseURL": "https://api.x.ai/v1",
        "models": {
          "default": [
            "grok-3-mini-latest"
          ],
          "fetch": true
        },
        "titleConvo": true,
        "titleMethod": "completion",
        "titleModel": "grok-2-latest",
        "summarize": false,
        "summaryModel": "grok-2-latest",
        "forcePrompt": false,
        "modelDisplayLabel": "Grok"
      },
      {
        "name": "Univ Ollama",
        "apiKey": "ollama-key-*********",
        "baseURL": "https://***************/v1",
        "models": {
          "default": [
            "mix_77/gemma3-qat-tools:27b",
            "MHKetbi/sbintuitions-sarashina2.2-3b-instruct-v0.1:Q8_0",
            "hf.co/alfredplpl/llm-jp-3-13b-instruct-gguf:Q8_0",
            "gemma3:27b-it-qat"
          ],
          "fetch": true
        },
        "titleConvo": true,
        "titleMethod": "completion",
        "titleModel": "MHKetbi/sbintuitions-sarashina2.2-3b-instruct-v0.1:Q8_0",
        "summarize": true,
        "summaryModel": "MHKetbi/sbintuitions-sarashina2.2-3b-instruct-v0.1:Q8_0",
        "modelDisplayLabel": "Ollama"
      }
    ]
  },
  "mcpServers": {
    "playwright": {
      "url": "http://mcp_container:8931/sse"
    }
  }
}
2025-07-26 14:45:46 warn: The `CHECK_BALANCE` environment variable is deprecated. Please use the `balance` field in the `librechat.yaml` config file instead.
More info: https://librechat.ai/docs/configuration/librechat_yaml/object_structure/balance#overview
2025-07-26 14:45:46 info: RAG API is running and reachable at http://rag_api:8000.
2025-07-26 14:45:46 info: Migrating 'USER' role from old schema: found 'PROMPTS' at top level
2025-07-26 14:45:46 info: Migrating 'USER' role from old schema: found 'BOOKMARKS' at top level
2025-07-26 14:45:46 info: Migrating 'USER' role from old schema: found 'AGENTS' at top level
2025-07-26 14:45:46 info: Migrating 'USER' role from old schema: found 'MULTI_CONVO' at top level
2025-07-26 14:45:46 info: Migrating 'USER' role from old schema: found 'TEMPORARY_CHAT' at top level
2025-07-26 14:45:46 info: Migrating 'USER' role from old schema: found 'RUN_CODE' at top level
2025-07-26 14:45:46 info: Unsetting old schema fields for 'USER' role: PROMPTS, BOOKMARKS, AGENTS, MULTI_CONVO, TEMPORARY_CHAT, RUN_CODE
2025-07-26 14:45:46 info: Updated role 'USER' and removed old schema fields
2025-07-26 14:45:46 info: Updated 'USER' role permissions
2025-07-26 14:45:46 info: Migrating 'ADMIN' role from old schema: found 'PROMPTS' at top level
2025-07-26 14:45:46 info: Migrating 'ADMIN' role from old schema: found 'BOOKMARKS' at top level
2025-07-26 14:45:46 info: Migrating 'ADMIN' role from old schema: found 'AGENTS' at top level
2025-07-26 14:45:46 info: Migrating 'ADMIN' role from old schema: found 'MULTI_CONVO' at top level
2025-07-26 14:45:46 info: Migrating 'ADMIN' role from old schema: found 'TEMPORARY_CHAT' at top level
2025-07-26 14:45:47 info: Migrating 'ADMIN' role from old schema: found 'RUN_CODE' at top level
2025-07-26 14:45:47 info: Unsetting old schema fields for 'ADMIN' role: PROMPTS, BOOKMARKS, AGENTS, MULTI_CONVO, TEMPORARY_CHAT, RUN_CODE
2025-07-26 14:45:47 info: Updated role 'ADMIN' and removed old schema fields
2025-07-26 14:45:47 info: Updated 'ADMIN' role permissions
2025-07-26 14:45:47 info: Turnstile is DISABLED (no siteKey provided).
2025-07-26 14:45:47 info:
Outdated Config version: 1.1.5
Latest version: 1.2.8

      Check out the Config changelogs for the latest options and features added.

      https://www.librechat.ai/changelog


2025-07-26 14:45:47 info: Configuring social logins...
2025-07-26 14:45:47 info: Server listening on all interfaces at port 3080. Use http://localhost:3080 to access it
2025-07-26 14:45:47 info: Initializing MCP servers...
2025-07-26 14:45:47 info: [indexSync] Messages are fully synced: 12832/12832
2025-07-26 14:45:47 info: [indexSync] Conversations are fully synced: 2635/2635
2025-07-26 14:45:47 info: [MCP][playwright] Access token missing
2025-07-26 14:45:47 info: [MCP][playwright] Access token missing and no refresh token available
2025-07-26 14:45:47 info: [MCP][playwright] Setting up OAuth event listener
2025-07-26 14:45:47 info: [MCP][playwright] Creating SSE transport: http://mcp_container:8931/sse
2025-07-26 14:45:47 info: [MCP][playwright] Instructions not included (serverInstructions not configured)
2025-07-26 14:45:47 info: [MCP][playwright] Capabilities: {"tools":{}}
2025-07-26 14:45:47 info: [MCP][playwright] Available tools: browser_close, browser_resize, browser_console_messages, browser_handle_dialog, browser_evaluate, browser_file_upload, browser_install, browser_press_key, browser_type, browser_navigate, browser_navigate_back, browser_navigate_forward, browser_network_requests, browser_take_screenshot, browser_snapshot, browser_click, browser_drag, browser_hover, browser_select_option, browser_tab_list, browser_tab_new, browser_tab_select, browser_tab_close, browser_wait_for
2025-07-26 14:45:47 info: [MCP] Initialized 1/1 app-level server(s)
2025-07-26 14:45:47 info: [MCP][playwright] ✓ Initialized
2025-07-26 14:45:47 info: [MCP] All app-level servers initialized successfully
2025-07-26 14:45:47 info: MCP servers initialized successfully
2025-07-26 14:45:57 error: Failed to fetch models from Univ Ollama API The server responded with status 502: Request failed with status code 502
2025-07-26 14:47:41 error: [api/server/controllers/agents/client.js #sendCompletion] Operation aborted 422 {"error":["Input should be 'tool'","Field required","Input should be 'system'","Input should be 'assistant'","Input should be a valid string"]}
2025-07-26 14:47:41 error: [api/server/controllers/agents/client.js #sendCompletion] Unhandled error type 422 {"error":["Input should be 'tool'","Field required","Input should be 'system'","Input should be 'assistant'","Input should be a valid string"]}
2025-07-26 14:48:26 error: Title error: Title generation timeout
2025-07-26 14:48:26 error: [api/server/controllers/agents/client.js #titleConvo] Error Aborted
2025-07-26 14:50:49 error: [api/server/controllers/agents/client.js #sendCompletion] Operation aborted 404 404 page not found


Troubleshooting URL: https://js.langchain.com/docs/troubleshooting/errors/MODEL_NOT_FOUND/

2025-07-26 14:50:49 error: [api/server/controllers/agents/client.js #sendCompletion] Unhandled error type 404 404 page not found


Troubleshooting URL: https://js.langchain.com/docs/troubleshooting/errors/MODEL_NOT_FOUND/

2025-07-26 14:50:59 error: [MCP][playwright] Transport error: SSE error: undefined
2025-07-26 14:50:59 info: [MCP][playwright] Reconnecting 1/3 (delay: 2000ms)
2025-07-26 14:51:01 info: [MCP][playwright] SSE transport closed
2025-07-26 14:51:01 info: [MCP][playwright] Creating SSE transport: http://mcp_container:8931/sse
2025-07-26 14:52:02 error: [api/server/controllers/agents/client.js #sendCompletion] Operation aborted 404 Unknown model: /AI21_Jamba_1.5_Large

Troubleshooting URL: https://js.langchain.com/docs/troubleshooting/errors/MODEL_NOT_FOUND/

2025-07-26 14:52:02 error: [api/server/controllers/agents/client.js #sendCompletion] Unhandled error type 404 Unknown model: /AI21_Jamba_1.5_Large

Troubleshooting URL: https://js.langchain.com/docs/troubleshooting/errors/MODEL_NOT_FOUND/

2025-07-26 14:52:40 error: [api/server/controllers/agents/client.js #sendCompletion] Operation aborted 422 {"error":["Input should be 'tool'","Field required","Input should be 'system'","Input should be 'assistant'","Input should be a valid string"]}
2025-07-26 14:52:40 error: [api/server/controllers/agents/client.js #sendCompletion] Unhandled error type 422 {"error":["Input should be 'tool'","Field required","Input should be 'system'","Input should be 'assistant'","Input should be a valid string"]}

Screenshots

No response

Code of Conduct

• I agree to follow this project's Code of Conduct

[danny-avila]

Try updating to latest and try again.

Updating Instructions (docker):
https://www.librechat.ai/docs/local/docker#update-librechat

[onestardao]

Yo, I ran into the exact same pattern before, and just so you know — this matches No.15 on our internal diagnostic map:

Misleading success from CLI/API test, but full app call fails due to hidden environment or token scope mismatches.

The kicker is:

The curl call works

The config.yaml seems fine

But the model silently dies inside your app runtime due to token permission mismatch or schema fallback.
It’s a super deceptive bug — especially when logs show 422/404 but don't say why.

I've got a fix (fully tested, open-source, MIT license), and it’s already helped a few other teams avoid this trap.
If you're interested, happy to share the module or walk you through what’s breaking in the backend.

Also, my project has some endorsements (ex-Tesseract.js folks etc), so I’m not just some random config whisperer, promise.

Let me know if you want the patch, or wanna dig deeper.