TRAM REST API Reference

Base URL: http://localhost:8765 (configurable via TRAM_HOST/TRAM_PORT)

All responses are JSON unless noted. Errors return {"detail": "message"}.

Health & Meta

GET /api/health

Liveness probe. Returns 200 immediately if daemon is running.

{"status": "ok"}

GET /api/ready

Readiness probe. Returns 200 once startup is complete, 503 if DB or scheduler is unavailable.

{
  "status": "ready",
  "db": "ok",
  "db_engine": "sqlite",
  "db_path": "/data/tram.db",
  "scheduler": "running",
  "cluster": "manager · 3/3 workers",
  "pipelines_loaded": 3,
  "uptime": "2h 15m 30s"
}

cluster values: "manager · N/M workers" in manager mode, "standalone" in standalone mode.

GET /api/meta

Build and version information.

{"version": "1.4.0", "build_time": "2026-05-01T12:00:00+00:00", "python_version": "3.13.0"}

GET /api/plugins

All registered plugin keys by category.

{
  "sources": ["kafka", "webhook", "websocket", "..."],
  "sinks": ["kafka", "opensearch", "elasticsearch", "..."],
  "serializers": ["json", "csv", "xml", "avro", "parquet", "msgpack", "protobuf"],
  "transforms": ["rename", "cast", "filter", "..."]
}

GET /api/cluster/nodes

Worker pool status (manager mode) or standalone indicator.

Standalone:

{"mode": "standalone", "workers": []}

Manager mode:

{
  "mode": "manager",
  "workers": [
    {
      "url": "http://trishul-ram-worker-0...:8766",
      "ok": true,
      "active_runs": 0,
      "running_pipelines": [],
      "assigned_pipelines": ["snmp_ifmib_to_sftp"]
    }
  ]
}

GET /api/cluster/streams

Active stream placement and throughput summary.

{
  "mode": "manager",
  "streams": [
    {
      "pipeline_name": "prom-ingest",
      "placement_group_id": "prom-ingest-20260417-ab12",
      "status": "degraded",
      "target_count": "all",
      "started_at": "2026-04-17T15:00:00+00:00",
      "slot_count": 3,
      "active_slots": 2,
      "records_in": 7102,
      "records_out": 7102,
      "records_skipped": 0,
      "dlq_count": 0,
      "error_count": 1,
      "bytes_in": 4505600,
      "bytes_out": 4505600,
      "records_in_per_sec": 1420.4,
      "records_out_per_sec": 1420.4,
      "bytes_in_per_sec": 901120.0,
      "bytes_out_per_sec": 901120.0,
      "slots": [
        {
          "worker_index": 0,
          "worker_id": "tram-worker-0",
          "worker_url": "http://trishul-ram-worker-0.trishul-ram-worker.default.svc.cluster.local:8766",
          "run_id_prefix": "prom-ingest-20260417-ab12-w0",
          "current_run_id": "prom-ingest-20260417-ab12-w0-r1",
          "status": "running",
          "restart_count": 1,
          "stats": {
            "schedule_type": "stream",
            "timestamp": "2026-04-17T15:00:05+00:00",
            "uptime_seconds": 5.0,
            "records_in": 3551,
            "records_out": 3551,
            "records_skipped": 0,
            "dlq_count": 0,
            "error_count": 0,
            "bytes_in": 2252800,
            "bytes_out": 2252800,
            "errors_last_window": [],
            "stale": false,
            "records_in_per_sec": 710.2,
            "records_out_per_sec": 710.2,
            "bytes_in_per_sec": 450560.0,
            "bytes_out_per_sec": 450560.0
          }
        },
        {
          "worker_index": 1,
          "worker_id": "tram-worker-1",
          "worker_url": "http://trishul-ram-worker-1.trishul-ram-worker.default.svc.cluster.local:8766",
          "run_id_prefix": "prom-ingest-20260417-ab12-w1",
          "current_run_id": "prom-ingest-20260417-ab12-w1-r0",
          "status": "stale",
          "restart_count": 0,
          "stats": {
            "schedule_type": "stream",
            "timestamp": "2026-04-17T14:58:00+00:00",
            "uptime_seconds": 5.0,
            "records_in": 3551,
            "records_out": 3551,
            "records_skipped": 0,
            "dlq_count": 0,
            "error_count": 1,
            "bytes_in": 2252800,
            "bytes_out": 2252800,
            "errors_last_window": [
              "timeout talking to sink"
            ],
            "stale": true,
            "records_in_per_sec": 0.0,
            "records_out_per_sec": 0.0,
            "bytes_in_per_sec": 0.0,
            "bytes_out_per_sec": 0.0
          }
        }
      ]
    }
  ]
}

Pipelines

GET /api/pipelines

List all registered pipelines with live status.

[
  {
    "name": "pm-ingest",
    "enabled": true,
    "status": "running",
    "schedule_type": "interval",
    "last_run": "2026-03-03T12:00:00Z",
    "last_run_status": "success"
  }
]

POST /api/pipelines

Register a new pipeline. Body: raw YAML text (Content-Type: text/plain or application/yaml) or JSON with yaml_text field.

curl -X POST http://localhost:8765/api/pipelines \
  -H "Content-Type: text/plain" \
  --data-binary @my-pipeline.yaml

Response: 201 Created — pipeline state dict.

Auto-saves a pipeline version to SQLite and auto-starts if enabled: true and schedule is not manual.

GET /api/pipelines/{name}

Get pipeline config and live status.

GET /api/pipelines/{name}/placement

Per-slot placement view for an active multi-worker stream, or a synthetic single-slot view for an active standalone stream pipeline (v1.3.2+).

Returns 404 when: the pipeline has no active multi-worker placement (manager mode), or the stream is not yet reporting stats / has stopped (standalone mode), or the pipeline is not a stream type.

{
  "pipeline_name": "prom-ingest",
  "placement_group_id": "prom-ingest-20260417-ab12",
  "status": "running",
  "target_count": "all",
  "started_at": "2026-04-17T15:00:00+00:00",
  "slot_count": 1,
  "active_slots": 1,
  "records_in": 7102,
  "records_out": 7102,
  "records_skipped": 0,
  "dlq_count": 0,
  "error_count": 0,
  "bytes_in": 4505600,
  "bytes_out": 4505600,
  "records_in_per_sec": 710.2,
  "records_out_per_sec": 710.2,
  "bytes_in_per_sec": 450560.0,
  "bytes_out_per_sec": 450560.0,
  "slots": [
    {
      "worker_index": 0,
      "worker_id": "tram-worker-0",
      "worker_url": "http://trishul-ram-worker-0.trishul-ram-worker.default.svc.cluster.local:8766",
      "run_id_prefix": "prom-ingest-20260417-ab12-w0",
      "current_run_id": "prom-ingest-20260417-ab12-w0-r1",
      "status": "running",
      "restart_count": 0,
      "stats": {
        "schedule_type": "stream",
        "timestamp": "2026-04-17T15:00:05+00:00",
        "uptime_seconds": 10.0,
        "records_in": 7102,
        "records_out": 7102,
        "records_skipped": 0,
        "dlq_count": 0,
        "error_count": 0,
        "bytes_in": 4505600,
        "bytes_out": 4505600,
        "errors_last_window": [],
        "stale": false,
        "records_in_per_sec": 710.2,
        "records_out_per_sec": 710.2,
        "bytes_in_per_sec": 450560.0,
        "bytes_out_per_sec": 450560.0
      }
    }
  ]
}

PUT /api/pipelines/{name}

Update/replace a registered pipeline’s YAML config in-place (v1.0.4). Stops the pipeline, re-registers with the new config, and restarts it if enabled: true. Body: raw YAML text (Content-Type: application/yaml or text/plain).

curl -X PUT http://localhost:8765/api/pipelines/pm-ingest \
  -H "Content-Type: application/yaml" \
  --data-binary @pm-ingest-updated.yaml

DELETE /api/pipelines/{name}

Deregister pipeline (stops it first). Returns 204 No Content.

POST /api/pipelines/{name}/start

Start scheduling or stream execution.

POST /api/pipelines/{name}/stop

Stop pipeline gracefully.

POST /api/pipelines/{name}/run

Trigger one immediate batch run (not valid for stream pipelines).

{"name": "pm-ingest", "status": "triggered"}

Queued response (v1.4.0) — in manager+worker mode with no healthy workers and queued runs enabled (TRAM_QUEUE_MANUAL_RUNS=1, the default), the run is durably queued instead of failing: it survives manager restarts and is dispatched automatically when worker capacity returns. The response is 202 Accepted with the stable run_id and absolute TTL:

{"name": "pm-ingest", "status": "queued", "run_id": "…", "expires_at": "…"}

Flush runs (v1.4.0) — ?flush=true makes stateful transforms emit their open windows as partials (window_complete: false) and clear them from the saved state. The flag is not carried through the queue: a queued flush run executes as a normal run when capacity returns — re-issue ?flush=true once capacity is back to flush.

POST /api/pipelines/reload

Re-scan TRAM_PIPELINE_DIR, reload all YAML files.

{"reloaded": 3, "total": 3}

Pipeline Dry-Run (v1.0.7)

POST /api/pipelines/dry-run

Validate and parse a pipeline YAML without registering it. Instantiates all transforms and connectors to catch config errors early.

Body: raw YAML text (Content-Type: text/plain or application/yaml).

curl -X POST http://localhost:8765/api/pipelines/dry-run \
  -H "Content-Type: text/plain" \
  --data-binary @my-pipeline.yaml

Response 200 (valid):

{"valid": true, "issues": []}

Response 200 (invalid — always 200, check valid field):

{"valid": false, "issues": ["serializer_in: unknown type 'xtf'"]
}

Pipeline Templates (v1.1.0)

GET /api/templates

List all bundled pipeline YAML templates from the pipelines/ directory.

[
  {"name": "kafka-to-opensearch", "description": "...", "yaml": "pipeline:\n  name: ..."},
  {"name": "snmp-poll-ifmib-to-influxdb", "description": "...", "yaml": "..."}
]

Pipeline Alerts (v1.0.0 / UI v1.1.0)

Alert rules evaluate simpleeval expressions after every batch run and fire webhook or email actions.

GET /api/pipelines/{name}/alerts

List alert rules for a pipeline.

[
  {
    "name": "high-error-rate",
    "condition": "error_rate > 0.05",
    "action": "webhook",
    "webhook_url": "https://hooks.example.com/alert",
    "cooldown_seconds": 300
  }
]

POST /api/pipelines/{name}/alerts

Add a new alert rule.

{
  "name": "low-output",
  "condition": "records_out < 10",
  "action": "email",
  "email_to": "ops@example.com",
  "subject": "Low output on pm-ingest",
  "cooldown_seconds": 600
}

PUT /api/pipelines/{name}/alerts/{rule_name}

Update an existing alert rule. Body: same structure as POST.

DELETE /api/pipelines/{name}/alerts/{rule_name}

Delete an alert rule. Returns 204 No Content.

Alert condition variables: records_in, records_out, records_skipped, error_rate, status, failed, duration_seconds.

Pipeline Versioning (v0.5.0)

GET /api/pipelines/{name}/versions

List all saved versions for a pipeline (requires SQLite persistence).

[
  {"id": 2, "name": "pm-ingest", "version": 2, "created_at": "2026-03-03T12:05:00Z", "is_active": 1},
  {"id": 1, "name": "pm-ingest", "version": 1, "created_at": "2026-03-03T12:00:00Z", "is_active": 0}
]

POST /api/pipelines/{name}/rollback?version=N

Restore pipeline to a previously saved version. Stops if running, reloads config, restarts if enabled.

{
  "name": "pm-ingest",
  "status": "stopped",
  "rolled_back_to_version": 1
}

Runs

GET /api/runs

Run history. Query params:

Param Default Description
pipeline — Filter by pipeline name
limit 100 Max records to return
status — Filter: success | failed | aborted
offset 0 Pagination offset (v0.7.0)
from_dt — ISO8601 lower bound on started_at (v0.7.0)
format — Set to csv to get text/csv export (v1.0.0)

With SQLite/DB persistence, run history survives daemon restarts.

[
  {
    "run_id": "abc12345",
    "pipeline": "pm-ingest",
    "node_id": "tram-0",
    "status": "success",
    "started_at": "2026-03-03T12:00:00Z",
    "finished_at": "2026-03-03T12:00:05Z",
    "records_in": 1500,
    "records_out": 1487,
    "records_skipped": 13,
    "bytes_in": 2359296,
    "bytes_out": 2341888,
    "dlq_count": 0,
    "error": null,
    "errors": ["Records skipped — no sink wrote successfully (condition filtered all records)"]
  }
]

GET /api/runs/{run_id}

Get a single run result.

Internal Transform State (v1.4.0)

Durable per-pipeline state for stateful transforms (counter_delta, window_aggregate). In manager+worker mode the worker GETs the state at run start and PUTs it back only after a successful run (retries re-hydrate from the same in-run snapshot); in standalone mode the state lives in the local transform_state table. Requires TRAM_STATEFUL_TRANSFORMS=1 (default); 0 disables both the transforms and these endpoints (404). update()/delete() on the pipeline purge the row; a config-hash mismatch discards the stored state so the new transform identities start fresh.

GET /api/internal/transform-state/{pipeline}

Returns the persisted state blob and its config hash.

{"pipeline": "pm-counters", "state": {"counter_delta:0": {"…identity…": {"v": 1500, "t": 1789534920.0}}}, "config_sha256": "1c3036a4cf884027"}

PUT /api/internal/transform-state/{pipeline}

Replaces the state. Bodies over TRAM_STATE_MAX_BYTES (default 20 MiB) are rejected with 413.

Live Stats (v1.1.0)

GET /api/stats

Per-pipeline aggregated stats for the last hour (records in/out, error rate, avg duration). Used by the Live Metrics Dashboard.

[
  {
    "pipeline": "pm-ingest",
    "records_in": 45000,
    "records_out": 44823,
    "error_rate": 0.004,
    "avg_duration_seconds": 1.23,
    "run_count": 60
  }
]

Authentication (v1.0.0)

When TRAM_API_KEY is set (or apiKey in Helm values), all protected /api/* requests must include the key via the X-API-Key header:

curl -H "X-API-Key: mysecret" http://localhost:8765/api/pipelines

The ?api_key= query param is removed — keys in URLs end up in access/proxy logs and browser history.

Exempt paths (always unauthenticated): /api/health, /api/ready, /agent/health, /metrics, /, /api/auth/login, /favicon.ico, /docs, /redoc, /openapi.json, and the /webhooks/* and /ui prefixes.

Internal machine-to-machine surfaces (/api/internal/* on the manager, /agent/* on workers) honor TRAM_INTERNAL_AUTH_MODE:

An invalid TRAM_INTERNAL_AUTH_MODE value is logged at WARNING and falls back to warn.

Returns 401 Unauthorized when the key is missing or wrong.

Browser Authentication (v1.0.8)

Set TRAM_AUTH_USERS (comma-separated username:password pairs) to bootstrap browser login. If TRAM_DB_URL is configured, changed passwords are stored in the user_passwords table and continue to work even after TRAM_AUTH_USERS is removed. Machine clients continue to use X-API-Key; browser users get 8-hour session tokens.

POST /api/auth/login

Authenticate with username and password. Verification prefers the DB-stored hash when present; otherwise it falls back to TRAM_AUTH_USERS.

curl -X POST http://localhost:8765/api/auth/login \
  -H "Content-Type: application/json" \
  -d '{"username": "admin", "password": "secret"}'

Response:

{"token": "eyJ...", "username": "admin"}

Use the token as Authorization: Bearer <token> on subsequent requests.

GET /api/auth/me

Returns the currently authenticated user from the Bearer token. Returns 401 if unauthenticated. Works for both env-bootstrapped and DB-backed browser auth.

{"username": "admin"}

POST /api/auth/change-password (v1.1.0)

Change the password for the currently authenticated user. Requires TRAM_DB_URL. New passwords are stored as scrypt$<salt>$<digest> hashes in the user_passwords DB table, persist across restarts, and override TRAM_AUTH_USERS for that user.

curl -X POST http://localhost:8765/api/auth/change-password \
  -H "Authorization: Bearer <token>" \
  -H "Content-Type: application/json" \
  -d '{"current_password": "old", "new_password": "new-secure-pass"}'

Response 200:

{"ok": true, "username": "admin"}

Connector Test (v1.1.0)

POST /api/connectors/test

Test connectivity for a single connector config. Returns whether the connection succeeded and any error message.

curl -X POST http://localhost:8765/api/connectors/test \
  -H "Content-Type: application/json" \
  -d '{"type": "kafka", "bootstrap_servers": "kafka:9092", "topic": "test"}'

Response:

{"ok": true, "latency_ms": 12, "error": null}

On failure: {"ok": false, "latency_ms": null, "error": "Connection refused"}

POST /api/connectors/test-pipeline

Test all source and sink connectors declared in a pipeline YAML. Returns per-connector results.

curl -X POST http://localhost:8765/api/connectors/test-pipeline \
  -H "Content-Type: text/plain" \
  --data-binary @my-pipeline.yaml

Response:

{
  "source": {"type": "sftp", "ok": true, "latency_ms": 45, "error": null},
  "sinks": [
    {"type": "kafka", "ok": false, "latency_ms": null, "error": "Connection refused"}
  ]
}

AI Assist (v1.1.0)

GET /api/ai/status

Returns AI configuration and availability.

{"available": true, "provider": "anthropic", "model": "claude-sonnet-4-6"}

Returns {"available": false} when TRAM_AI_API_KEY is not set.

POST /api/ai/suggest

Generate or explain a pipeline YAML using AI.

Field Description
mode "generate" — create a new pipeline from a description; "explain" — explain existing YAML
prompt Natural language description (for generate) or question (for explain)
yaml Existing pipeline YAML (required for explain mode) 
# Generate
curl -X POST http://localhost:8765/api/ai/suggest \
  -H "Content-Type: application/json" \
  -d '{"mode": "generate", "prompt": "Poll SNMP IF-MIB every 60s and write to InfluxDB"}'

# Explain
curl -X POST http://localhost:8765/api/ai/suggest \
  -H "Content-Type: application/json" \
  -d '{"mode": "explain", "prompt": "What does this do?", "yaml": "pipeline:\n  name: ..."}'

Response:

{"result": "pipeline:\n  name: snmp-to-influxdb\n  ..."}

Configure via env vars:

Env Var Description
TRAM_AI_API_KEY API key for the AI provider
TRAM_AI_PROVIDER anthropic, openai, or bedrock (default: anthropic)
TRAM_AI_MODEL Model name (defaults: claude-haiku-4-5-20251001 for Anthropic, gpt-4o-mini for OpenAI, us.anthropic.claude-sonnet-4-6 for Bedrock)
TRAM_AI_BASE_URL Custom base URL (e.g. for Ollama or Azure OpenAI)

Webhooks (v0.5.0)

POST /webhooks/{path}

Forward a raw HTTP POST body to a registered webhook source pipeline.

curl -X POST http://localhost:8765/webhooks/my-events \
  -H "Content-Type: application/json" \
  -d '{"ne_id": "node-1", "severity": 2}'

To enable: add a pipeline with source.type: webhook and source.path: my-events.

Metrics (v0.5.0)

GET /metrics

Prometheus metrics in text exposition format (text/plain; version=0.0.4).

Returns 503 with JSON error if prometheus_client is not installed.

# HELP tram_records_in_total Total records read from source
# TYPE tram_records_in_total counter
tram_records_in_total{pipeline="pm-ingest"} 45000.0

# HELP tram_records_out_total Total records written to sink
# TYPE tram_records_out_total counter
tram_records_out_total{pipeline="pm-ingest"} 44823.0

# HELP tram_records_skipped_total Total records skipped
# TYPE tram_records_skipped_total counter
tram_records_skipped_total{pipeline="pm-ingest"} 177.0

# HELP tram_errors_total Total processing errors
# TYPE tram_errors_total counter
tram_errors_total{pipeline="pm-ingest"} 0.0

# HELP tram_chunk_duration_seconds Time spent processing one chunk
# TYPE tram_chunk_duration_seconds histogram
tram_chunk_duration_seconds_bucket{le="0.01",pipeline="pm-ingest"} 120.0
...

SNMP MIBs (v1.0.3)

Manages compiled pysnmp MIB .py files in TRAM_MIB_DIR (default /mibs). Standard MIBs (IF-MIB, ENTITY-MIB, HOST-RESOURCES-MIB, IP-MIB, TCP-MIB, UDP-MIB, IANAifType-MIB) are pre-compiled in the Docker image.

GET /api/mibs

List all compiled MIB modules in TRAM_MIB_DIR.

[
  {"name": "IF-MIB", "file": "IF-MIB.py", "size_bytes": 14823},
  {"name": "ENTITY-MIB", "file": "ENTITY-MIB.py", "size_bytes": 22104}
]

POST /api/mibs/upload

Upload a raw .mib text file and compile it. Requires tram[mib]; returns 501 if not installed.

curl -X POST http://localhost:8765/api/mibs/upload \
  -F "file=@MY-CUSTOM-MIB.mib"

Response:

{"compiled": ["MY-CUSTOM-MIB"], "mib_dir": "/mibs", "results": {"MY-CUSTOM-MIB": "compiled"}}

POST /api/mibs/download

Download and compile MIB modules by name from mibs.pysnmp.com. Requires internet access and tram[mib].

curl -X POST http://localhost:8765/api/mibs/download \
  -H "Content-Type: application/json" \
  -d '{"names": ["CISCO-ENTITY-FRU-CONTROL-MIB", "CISCO-TC-MIB"]}'

DELETE /api/mibs/{name}

Delete a compiled MIB module from TRAM_MIB_DIR.

curl -X DELETE http://localhost:8765/api/mibs/MY-CUSTOM-MIB

Schemas (v1.0.3)

Manages serialization schema files (.proto, .avsc, .json, .xsd, .yaml, .yml) in TRAM_SCHEMA_DIR (default /schemas). No compilation — files are stored as-is. Pipeline executors compile or read them at run time.

GET /api/schemas

List all schema files under TRAM_SCHEMA_DIR recursively.

[
  {
    "path": "cisco/GenericRecord.proto",
    "type": "protobuf",
    "size_bytes": 3421,
    "schema_file": "/schemas/cisco/GenericRecord.proto"
  },
  {
    "path": "events.avsc",
    "type": "avro",
    "size_bytes": 892,
    "schema_file": "/schemas/events.avsc"
  }
]

schema_file is the absolute path ready to paste into a pipeline schema_file: field.

type is inferred from the extension: protobuf, avro, json, xml, yaml, other.

GET /api/schemas/{filepath}

Return the raw text content of a schema file. filepath is relative to TRAM_SCHEMA_DIR.

curl http://localhost:8765/api/schemas/cisco/GenericRecord.proto

Returns 404 if not found, 400 if the path escapes TRAM_SCHEMA_DIR.

POST /api/schemas/upload

Upload a schema file. Accepts .proto, .avsc, .json, .xsd, .yaml, .yml. Returns 400 for other extensions.

Query param Description
subdir Optional subdirectory within TRAM_SCHEMA_DIR (e.g. cisco). Must not contain ...

Upload all Cisco EMS proto files to a shared subdirectory:

for f in *.proto; do
  curl -F "file=@$f" \
    "http://localhost:8765/api/schemas/upload?subdir=cisco"
done

Response:

{
  "path": "cisco/GenericRecord.proto",
  "type": "protobuf",
  "size_bytes": 3421,
  "schema_file": "/schemas/cisco/GenericRecord.proto",
  "schema_dir": "/schemas"
}

DELETE /api/schemas/{filepath}

Delete a schema file. filepath is relative to TRAM_SCHEMA_DIR.

curl -X DELETE http://localhost:8765/api/schemas/cisco/GenericRecord.proto

Returns 404 if not found, 400 on path-traversal attempt.

Schema Registry Proxy (v1.0.4)

Transparent reverse proxy to an external Confluent-compatible schema registry (e.g. Confluent Schema Registry, Apicurio Registry). Enabled by setting TRAM_SCHEMA_REGISTRY_URL.

All HTTP methods (GET, POST, PUT, DELETE, PATCH) are proxied. Request headers, query params, and body are forwarded as-is. This lets UI tools and serializer clients reach the external registry through a single origin (TRAM).

# List subjects
curl http://localhost:8765/api/schemas/registry/subjects

# Get latest schema for a subject
curl http://localhost:8765/api/schemas/registry/subjects/device-event-value/versions/latest

# Register a new schema version
curl -X POST http://localhost:8765/api/schemas/registry/subjects/device-event-value/versions \
  -H "Content-Type: application/vnd.schemaregistry.v1+json" \
  -d '{"schema": "{\"type\":\"record\",\"name\":\"DeviceEvent\",\"fields\":[]}"}'

Returns 503 Service Unavailable when TRAM_SCHEMA_REGISTRY_URL is not set. Returns 502 Bad Gateway when the upstream registry is unreachable.

Configuration:

Env Var Description
TRAM_SCHEMA_REGISTRY_URL Base URL of the external registry (e.g. http://schema-registry:8081)

Serializer auto-fallback — when TRAM_SCHEMA_REGISTRY_URL is set, Avro and Protobuf serializers automatically use it as their registry URL without requiring schema_registry_url: in pipeline YAML. Pipeline-level schema_registry_url: overrides the env default per-pipeline.

Daemon

GET /api/daemon/status

Scheduler state, active streams, next scheduled runs.

POST /api/daemon/stop

Graceful shutdown.

Error Responses

Code Meaning
400 Invalid pipeline YAML or config
401 Missing/invalid X-API-Key header (the legacy ?api_key= query param was removed in v1.4.0); or missing/invalid Authorization: Bearer for webhook secret
404 Pipeline, run, or webhook path not found
409 Pipeline already registered
422 Pydantic validation error
429 Rate limit exceeded (v1.0.0) — retry after the TRAM_RATE_LIMIT_WINDOW window resets
503 DB unavailable (readiness check); prometheus_client not installed (/metrics)