Generic webhook

Send events from any language or tool with a single HTTP POST — no SDK or library required.

Endpoint

text

POST https://api.obsivara.com/ingest/webhook/generic

The n8n alias (/ingest/webhook/n8n) accepts the same payload — use either. Every request must include Authorization: Bearer obs_live_YOUR_KEY andContent-Type: application/json.

Full example

bash

curl -X POST https://api.obsivara.com/ingest/webhook/generic \
  -H "Authorization: Bearer obs_live_YOUR_KEY" \
  -H "Content-Type: application/json" \
  -d '{
    "event": "run_started",
    "run_id": "my-run-001",
    "asset_id": "my-agent",
    "name": "First run",
    "input": "What is the weather today?"
  }'

Expected 202 response:

json

{ "source": "generic", "accepted": 1, "rejected": 0, "errors": [] }

Event types

The event field (snake_case) must be one of three values. Unknown values are stored as error rows — spell them exactly.

run_started

Marks the beginning of a workflow run.

json

{
  "event": "run_started",
  "run_id": "my-run-001",
  "asset_id": "my-agent",
  "name": "Support reply",
  "input": "Where is my order?"
}

Field	Type	Required	Description
`event`	string	Yes	Exactly `"run_started"`
`run_id`	string	Yes	Unique ID for this run. Any string — non-UUID values are auto-hashed to a stable UUID server-side.
`asset_id`	string	Yes	The agent, workflow, or prompt you are monitoring.
`name`	string	No	Human-readable label for the run.
`input`	string	No	The input text or prompt that triggered this run.

llm_call

Records a model call within a run.

json

{
  "event": "llm_call",
  "run_id": "my-run-001",
  "provider": "openai",
  "model": "gpt-4o",
  "prompt_tokens": 312,
  "completion_tokens": 87,
  "total_tokens": 399,
  "cost_usd": 0.0048,
  "latency_ms": 1240
}

Field	Type	Required	Description
`event`	string	Yes	Exactly `"llm_call"`
`run_id`	string	Yes	Must match the `run_id` from `run_started`.
`provider`	string	No	e.g. `"openai"`, `"anthropic"`, `"google"`
`model`	string	No	e.g. `"gpt-4o"`, `"claude-3-5-sonnet"`
`prompt_tokens`	number	No	Input token count.
`completion_tokens`	number	No	Output token count.
`total_tokens`	number	No	Sum of prompt + completion tokens.
`cost_usd`	number	No	Cost of this call in US dollars.
`latency_ms`	number	No	Time-to-first-response or total call duration in milliseconds.

run_finished

Marks the end of a run.

json

{
  "event": "run_finished",
  "run_id": "my-run-001",
  "outcome": "success",
  "duration_ms": 3500,
  "total_cost_usd": 0.0048
}

Field	Type	Required	Description
`event`	string	Yes	Exactly `"run_finished"`
`run_id`	string	Yes	Must match the `run_id` from `run_started`.
`outcome`	string	No	`"success"` or `"error"`
`duration_ms`	number	No	Total run duration in milliseconds.
`total_cost_usd`	number	No	Total cost of all LLM calls in the run.

Asset attribution

Obsivara resolves asset_id in this order (first non-empty wins):

asset_id field in the JSON body
?asset_id= query parameter
X-Obsivara-Asset-Id request header
"unknown" (fallback)

The run_id can be any string — Obsivara maps non-UUID values to a stable UUID automatically, so you can use your own IDs (order numbers, session IDs, etc.) without generating UUIDs.

Unknown event values are accepted (202) but stored as error rows. Spell event names exactly:run_started, llm_call, run_finished.

n8n alias

The endpoint /ingest/webhook/n8n is an alias for /ingest/webhook/generic — it accepts the same payload. See n8n integration → for a step-by-step guide with n8n expressions.

Next steps

Verify it worked — confirm your events appear in the dashboard
Event schema reference — all event types and optional fields for the batch API
Errors & troubleshooting — resolve 401, 403, 400 responses