{
  "checker": {
    "details": {
      "checks": [
        {
          "check": "install",
          "details": {
            "argv": [
              "<host-path>",
              "--help"
            ],
            "blocker": "product invocation failed: {\"returncode\": 0, \"stdout\": \"Together CLI (v2.42.0)\\n\\n\\u256d\\u2500 Commands \\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u256e\\n\\u2502    batches            Submit and manage batch jobs                           \\u2502\\n\\u2502    clusters           Create and manage GPU clusters                         \\u2502\\n\\u2502    endpoints          Deploy and manage dedicated endpoints                  \\u2502\\n\\u2502    evals              Run and manage model evaluations                       \\u2502\\n\\u2502    files              Upload and manage files                                \\u2502\\n\\u2502    ft, fine-tuning    Create and manage fine-tuning jobs                     \\u2502\\n\\u2502    models             List and upload models                                 \\u2502\\n\\u2502    projects           List projects you can access                           \\u2502\\n\\u2502    telemetry          Configure CLI telemetry                                \\u2502\\n\\u2502    training           Inspect training resources and prepare models for      \\u2502\\n\\u2502                       inference                                              \\u2502\\n\\u2502    whoami             Show the current user                                  \\u2502\\n\\u2502    login              Logs into your account or creates a new one            \\u2502\\n\\u2502    logout             Logs out of your account                               \\u2502\\n\\u2502    switch             Switches between different scopes                      \\u2502\\n\\u2570\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u256f\\n\\u256d\\u2500 Beta Commands \\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u256e\\n\\u2502    beta endpoints    Deploy and manage dedicated inference endpoints         \\u2502\\n\\u2502    beta jig          Container deployment                                    \\u2502\\n\\u2502    beta models       Register and manage models for dedicated inference      \\u2502\\n\\u2570\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u256f\\n\\u256d\\u2500 Global Options \\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u256e\\n\\u2502    --debug                    Print HTTP request/response details to stderr  \\u2502\\n\\u2502                               and write a full log, including bodies, to a   \\u2502\\n\\u2502                               temp file                                      \\u2502\\n\\u2502    --non-interactive          Disable interactive prompts                    \\u2502\\n\\u2502    --project          string  Together project ID. Defaults to               \\u2502\\n\\u2502                               TOGETHER_PROJECT_ID. If omitted, read-only     \\u2502\\n\\u2502                               commands use the project associated with the   \\u2502\\n\\u2502                               API key; mutating commands may prompt for      \\u2502\\n\\u2502                               confirmation or require an explicit project in \\u2502\\n\\u2502                               JSON or non-interactive mode.                  \\u2502\\n\\u2502    --no-json                  Output the response as JSON. Defaults to on    \\u2502\\n\\u2502                               when an AI agent is detected; pass --no-json   \\u2502\\n\\u2502                               to force text output.                          \\u2502\\n\\u2502    -h, --help                 Display this message and exit                  \\u2502\\n\\u2502    --version                  Display application version                    \\u2502\\n\\u2570\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u2500\\u256f\\n\\nExamples:\\n- Fine tune a model for your dataset:\\n  tg ft create --model Qwen/Qwen2-1.5B --training-file ./my-dataset.jsonl --lora\\n\\n- Deploy a model to a dedicated endpoint:\\n  Create an endpoint and deploy the model to it.\\n  After the endpoint is created, you can run inference against it with the fully\\nqualified endpoint name.\\n  tg beta endpoints deploy Qwen/Qwen2.5-7B --endpoint my-custom-endpoint-name\\n\\n- Upload an external model to Together:\\n  First create a model record and reference compatible base model.\\n  You can find model IDs of the supported base models with `tg beta models \\npublic`.\\n  $ tg beta models create --name my-model --base-model ml_xxxxxxxxxxxx\\n\\n  Upload from a local file/directory to the model record created above.\\n  $ tg beta models upload ml_yyyyyyyyyyy ./path/to/my-model\\n\\n\", \"stderr\": \"\", \"needs_credentials\": false, \"needs_browser\": false}",
            "kind": "cli",
            "needs_browser": false,
            "needs_credentials": false,
            "returncode": 0,
            "stderr": "",
            "stdout": "Together CLI (v2.42.0)\n\n\u256d\u2500 Commands \u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u2500\u256e\n\u2502    batches            Submit and manage batch jobs                           \u2502\n\u2502    clusters           Create and manage GPU clusters                         \u2502\n\u2502    endpoints          Deploy and manage dedicated endpoints                  \u2502\n\u2502    evals              Run and manage model evaluations                       \u2502\n\u2502    files              Upload and manage files                                \u2502\n\u2502    ft, fine-tuning    Create and manage fine-tuning jobs                     \u2502\n\u2502    models             List and upload models                                 \u2502\n\u2502    projects           List projects you can access                           \u2502\n\u2502    telemetry          Configure CLI telemetry                                \u2502\n[excerpt truncated]",
            "working": false
          },
          "optional": false,
          "status": "FAIL"
        },
        {
          "check": "credentials_error",
          "details": {
            "argv": [
              "<host-path>",
              "models",
              "list"
            ],
            "blocker": "product did not produce a clear credential error: {\"returncode\": 1, \"stdout\": \"Error: Not logged in. Run `tg login`.\\n\", \"stderr\": \"\", \"needs_credentials\": false, \"needs_browser\": false}",
            "kind": "cli",
            "needs_browser": false,
            "needs_credentials": false,
            "returncode": 1,
            "stderr": "",
            "stdout": "Error: Not logged in. Run `tg login`.",
            "working": false
          },
          "optional": false,
          "status": "FAIL"
        }
      ],
      "scope": "partial checks",
      "working": false
    },
    "status": "FAIL"
  },
  "deviations": [
    "python3-venv was not installed in the guest; ran `sudo apt-get update` and `sudo apt-get install python3.12-venv` (prerequisite install, not a repo defect).",
    "Did not run `tg login` or obtain any credentials; the task forbids obtaining credentials.",
    "Did not run the full CLI operation set; `tg files check` was also tried locally against a sample JSONL in /tmp/tgprobe and also returned 'Not logged in', so it is not a credential-free operation."
  ],
  "fault_owner": "prerequisite",
  "main_blocker": "No Together API credentials in the guest: `Error: Not logged in. Run `tg login`.` (rc=1).",
  "operation": "Install the SDK/CLI from the pinned checkout (README 'Installation' via pip), then run README 'Usage \u2013 CLI' > 'Models' > `tg models list` (the documented operation) to confirm the CLI reaches the Together API.",
  "operation_scope": "Install and CLI entry point verified (tg --help). The operation itself was not exercised: the guest has no Together API credentials (no TOGETHER_API_KEY, no `tg login` session), so every API-backed command, including `tg models list` and `tg files check`, exits 1 with 'Error: Not logged in. Run `tg login`.' Unmet prerequisite: a Together API key in the guest.",
  "outcome": "blocked",
  "steps": [
    "python3 -m venv <host-path> (needed apt python3.12-venv; installed via sudo apt-get)",
    "<host-path> install -e <host-path> (README 'Installation')",
    "<host-path> --version -> 2.42.0",
    "<host-path> --help (install check)",
    "cd /tmp && <host-path> models list -> rc=1 'Error: Not logged in. Run `tg login`.'"
  ],
  "stuck_step": "README 'Usage \u2013 CLI' > 'Models': `tg models list`",
  "suggested_fix": "Provide a Together API key to the guest via TOGETHER_API_KEY (or `tg login` with a key supplied by the operator), then rerun `tg models list` as the probe. Not a repo defect; no source was changed."
}