package v1.public.events.litellm_gateway

Get desktop application:
View/edit binary Protocol Buffers messages

Used in: LiteLlmRequestCompleted

• string api_key_hash = 1

  SHA-256 hash of the API key used for the request.

• optional int64 account_id = 2

  dbt Cloud account ID (from team_id on the virtual key).

• optional int64 user_id = 3

  dbt Cloud user ID (from key metadata).

• optional string account_identifier = 4

  Global dbt Cloud account identifier slug (from /api/v2/accounts/{id}/).

Used in: LiteLlmRequestCompleted

• bool is_billable_request = 1

  Deprecated: billability should be determined downstream from event_type, deployment_mode, and model metadata. Always sent as false.

• string event_subtype = 2

  Event subtype for analytics: "completion", "websocket_lifecycle".

Used in: LiteLlmRequestCompleted

• double response_cost_usd = 1

  Cost in USD as reported by LiteLLM.

• string cost_status = 2

  How the cost was determined: "calculated", "non_billable", "missing_usage", "pricing_failed".

• optional string cost_reason = 3

  Explanation when cost_status is not "calculated" (e.g. failure debug info).

Used in: LiteLlmRequestCompleted

• optional string error_class = 1

  Exception class name (e.g. "BadRequestError").

• optional string error_code = 2

  Provider error code.

• optional string error_message = 3

  Human-readable error description.

• optional string llm_provider = 4

  Provider that returned the error.

Used in: LiteLlmRequestCompleted

• double start_time = 1

  Unix timestamp when the request started.

• double end_time = 2

  Unix timestamp when the response completed.

• double response_time_seconds = 3

  Total response time in seconds.

Used in: LiteLlmRequestCompleted

• optional string api_base = 1

  Full provider endpoint URL.

• string model_group = 2

  Logical model group (e.g. "claude-opus-4-7").

• string model_id = 3

  Internal model deployment ID.

• string provider = 4

  LLM provider name (e.g. "bedrock", "openai").

• string requested_model = 5

  Model as requested by the caller (e.g. "bedrock/us.anthropic.claude-opus-4-7").

Used in: LiteLlmRequestCompleted

• string call_type = 1

  LiteLLM call type (e.g. "acompletion", "aresponses").

• string id = 2

  ChatCompletion response ID.

• string litellm_call_id = 3

  Internal LiteLLM correlation ID.

• string status = 4

  "success" or "failure".

• bool stream = 5

  Whether the request was streamed.

• string trace_id = 6

  Distributed trace ID for request correlation.

Emitted on every LLM request completion (success or failure) by the litellm gateway.

• optional events.vortex.VortexMessageEnrichment enrichment = 1

• optional LiteLlmActor actor = 2

• string event_type = 3

  "llm_request_completed" or "llm_request_failed".

• string installation_id = 4

  Pod/deployment identifier (e.g. "litellm-549bf484d9-cs58d").

• optional google.protobuf.Timestamp occurred_at = 5

• optional LiteLlmRequest request = 6

• optional LiteLlmModel model = 7

• optional LiteLlmLatency latency = 8

• optional LiteLlmUsage usage = 9

• optional LiteLlmError error = 10

  Populated only on failure events.

• optional LiteLlmWizardAttribution wizard_attribution = 11

  Wizard session/turn attribution from x-codex-turn-metadata.

• optional LiteLlmCost cost = 12

  Cost information derived from LiteLLM response_cost.

• optional LiteLlmClassification classification = 13

  Event classification for billing and analytics.

• string deployment_mode = 14

  "local" (BYOK sidecar) or "hosted" (Helm-deployed proxy).

Used in: LiteLlmRequestCompleted

• int64 prompt_tokens = 1

• int64 completion_tokens = 2

• int64 total_tokens = 3

• int64 cache_creation_input_tokens = 4

  Anthropic prompt caching: tokens written to cache.

• int64 cache_read_input_tokens = 5

  Anthropic prompt caching: tokens read from cache.

• int64 cached_tokens = 6

  OpenAI-style cached prompt tokens.

Used in: LiteLlmRequestCompleted

• string wizard_session_id = 1

  Wizard session ID from x-codex-turn-metadata.

• string wizard_turn_id = 2

  Wizard turn ID from x-codex-turn-metadata.

• optional string wizard_thread_source = 3

  Thread source (e.g. "cli", "vscode", "cloud").

• optional int64 wizard_turn_started_at_unix_ms = 4

  Unix timestamp in milliseconds when the turn started.

• bool is_wizard_internal = 5

  Whether the WIZARD_INTERNAL env var was set for this session.