Configuration Guide¶

This service uses a typed top-level TOML configuration file.

Overview¶

Configuration is split into two layers:

config/service.toml
service bootstrap settings
server parameters
worker parameters
input resolver settings
output sink settings
paths to processor-specific TOML files
specialized processor TOMLs
config/c2pa/*.toml
config/media-packager/*.toml

The service loads config/service.toml by default. Set SERVICE_CONFIG_PATH to point at a different file.

Service Config¶

Example:

[server]
bind_addr = "0.0.0.0:8080"
max_upload_bytes = 52428800
sync_timeout_ms = 120000
sync_poll_interval_ms = 250

[worker]
queue_capacity = 128
max_concurrent_batches = 4
max_files_parallel = 4

[input_resolvers.s3]
# profile = "my-aws-profile"
force_path_style = false
buckets = []

[output_sinks.local]
base_dir = "./.artifacts/output"

[output_sinks.s3]
# region = "eu-central-1"
# force_path_style = false
# buckets = ["processed-media", "archive-media"]
# base_prefix = "exports"

[processors.c2pa]
sdk_settings_path = "config/c2pa/default_c2pa_sdk_settings.toml"

[processors.media_packager]
pipeline_config_path = "config/media-packager/media_packager_pipeline.toml"

Sections¶

`server`¶

bind_addr: address and port for the HTTP server
max_upload_bytes: maximum HTTP request body size accepted by Axum
sync_timeout_ms: sync endpoint timeout
sync_poll_interval_ms: sync endpoint polling interval

`worker`¶

queue_capacity: job queue size
max_concurrent_batches: maximum batches processed in parallel
max_files_parallel: per-batch input parallelism, applied to every batch (requests cannot override it)

`input_resolvers.s3`¶

presence of this section enables S3 resolver bootstrap
profile: optional AWS shared config profile; when omitted the default credential chain is used
region: optional AWS region
endpoint_url: optional custom endpoint for MinIO/LocalStack
force_path_style: enable path-style addressing
buckets: allowed buckets and startup validation targets

`output_sinks.local`¶

base_dir: base directory for persisted outputs
naming: file-leaf naming mode for single-file outputs written to this sink; "derived" (default, appends _c2pa) or "in_place" (keep the input filename). Folder/publication outputs always use the input folder name.

`output_sinks.s3`¶

optional output sink configuration under the existing s3 feature
profile: optional AWS shared config profile
region: optional AWS region
endpoint_url: optional custom S3-compatible endpoint
force_path_style: enable path-style addressing
buckets: writable bucket allow-list for request-selected S3 outputs
base_prefix: optional shared prefix prepended to every uploaded object key
naming: file-leaf naming mode for single-file outputs, same values and default as output_sinks.local.naming

Behavior:

there is no default S3 output bucket
a request selects S3 output with output.type = "s3" and a bucket; output.type = "local" (or omitting output) targets the local sink instead
only configured buckets are accepted
configured output buckets are write-probed during startup
object keys are composed as base_prefix / output.prefix / output.name / <per-input-leaf>; there is no job/batch id segment. output.prefix = "/" (or omitted) writes directly under base_prefix; if no base_prefix is configured, keys start at the bucket root. Reusing a destination overwrites it.

`processors`¶

These sections reference specialized processor configuration files instead of inlining their internals:

processors.c2pa.sdk_settings_path
processors.media_packager.pipeline_config_path

Feature Flags¶

openapi: exposes /openapi.json
scalar-docs: extends openapi and exposes /docs
s3: enables S3-backed input resolvers and request-selectable S3 outputs

The synchronized public OpenAPI file is generated from the private service implementation using the documentation feature profile, which enables both openapi and s3.

If [input_resolvers.s3] or [output_sinks.s3] is configured without building the service with --features s3, startup fails.

Request Input Types¶

The current request model is provider-agnostic:

local_file_path
local_folder_path
remote_file_url
remote_folder_url

Examples:

{"type": "remote_file_url", "url": "https://example.com/video.mp4"}
{"type": "remote_file_url", "url": "s3://media-bucket/input/video.mp4"}
{"type": "remote_folder_url", "url": "s3://media-bucket/publication/42/"}

Request Output Types¶

The output section selects the destination and placement. Supported types:

local — writes under output_sinks.local.base_dir on the service host (dev/self-host). Omitting output is equivalent to {"type": "local"} with no fields.
s3 — writes to object storage; requires the s3 build feature and a configured bucket.

Both accept optional prefix and name. The file-leaf naming mode is not a request field; it is configured per output sink (output_sinks.*.naming, see above). See Output Selection for the full contract.