Antfly · Schema

EnrichmentConfig

Inline managed enrichment definition. Enrichments materialize generated artifacts before indexing and may target source rows or previously generated artifact streams.

Artificial IntelligenceSearchRetrievalVectorsHybrid

Properties

Name Type Description
name string Stable generated artifact name.
kind object
source object Required for kind=group. The covering membership index and ordinary unit artifacts are maintained from source row commits.
prior_context object Bounded earlier-segment context from an artifact sharing the same grouped source. Only valid for prompt-consuming asset producers.
field string Source field to read from the source document or source artifact payload.
template string Optional template for generated text input.
source_artifact_name string Existing artifact stream this enrichment consumes. Chunk enrichments may consume asset artifacts; embedding enrichments may consume chunk artifacts; asset enrichments may consume other asset artifacts
expected_dims integer Expected embedding dimension for embedding enrichments.
vector_space string Optional stable model/token-space identifier for embedding artifacts. When omitted on every source, Antfly requires the effective producers to be semantically equivalent. To combine intentionally comp
chunk_size integer Chunk size for chunk enrichments.
chunk_overlap integer Chunk overlap for chunk enrichments.
chunker object Chunker configuration for chunk enrichments. Cannot be combined with chunker_json.
chunker_json string Legacy serialized chunker configuration for chunk enrichments. Cannot be combined with chunker.
full_text_index boolean When true on a chunk or asset enrichment, route generated text into the table's default full-text index.
content_type string Produced asset content type for asset enrichments.
producer object Write-only producer configuration. Cannot be combined with producer_json or transcriber. Decision producers use type=decision and config={version, decider, questions}. A named-embedder similarity deci
producer_json string Write-only serialized producer configuration. For managed embedding enrichments Antfly stores a canonical semantic producer identity here; credentials and execution policy are excluded.
neighbor_context object Optional bounded sample of the document's graph neighbors appended to the producer input. Only valid on asset enrichments whose producer consumes rendered prompt text (generator, extractor, or decisio
execution object Non-semantic execution policy for this enrichment producer. This does not participate in generated artifact identity.
transcriber object Typed shorthand for a transcription asset enrichment. Only valid with kind=asset and without producer_json; Antfly expands it into a document_extraction producer whose audio route transcribes each rec
View JSON Schema on GitHub

JSON Schema

antfly-enrichment-config-schema.json Raw ↑
{
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "$id": "https://artifacts.apis.io/antfly/json-schema/antfly-enrichment-config-schema.json",
  "title": "EnrichmentConfig",
  "description": "Inline managed enrichment definition. Enrichments materialize generated artifacts before indexing and may target source rows or previously generated artifact streams.",
  "x-generated": "2026-11-10",
  "x-method": "derived",
  "x-generator": "derive-json-schema.py",
  "x-source": "openapi/antfly-index-management-api-openapi.yml#/components/schemas/EnrichmentConfig",
  "type": "object",
  "required": [
    "name",
    "kind"
  ],
  "properties": {
    "name": {
      "type": "string",
      "description": "Stable generated artifact name."
    },
    "kind": {
      "$ref": "#/$defs/EnrichmentKind"
    },
    "source": {
      "$ref": "#/$defs/GroupedArtifactSource",
      "description": "Required for kind=group. The covering membership index and ordinary unit artifacts are maintained from source row commits."
    },
    "prior_context": {
      "$ref": "#/$defs/PriorArtifactContext",
      "description": "Bounded earlier-segment context from an artifact sharing the same grouped source. Only valid for prompt-consuming asset producers."
    },
    "field": {
      "x-go-type-skip-optional-pointer": true,
      "type": "string",
      "description": "Source field to read from the source document or source artifact payload."
    },
    "template": {
      "x-go-type-skip-optional-pointer": true,
      "type": "string",
      "description": "Optional template for generated text input."
    },
    "source_artifact_name": {
      "x-go-type-skip-optional-pointer": true,
      "type": "string",
      "description": "Existing artifact stream this enrichment consumes. Chunk enrichments may consume asset artifacts; embedding enrichments may consume chunk artifacts; asset enrichments may consume other asset artifacts (the upstream asset's produced bytes become this producer's source, so field and template must be omitted and the producer must consume text: copy, generator, or extractor)."
    },
    "expected_dims": {
      "x-go-type-skip-optional-pointer": true,
      "type": "integer",
      "minimum": 0,
      "description": "Expected embedding dimension for embedding enrichments."
    },
    "vector_space": {
      "x-go-type-skip-optional-pointer": true,
      "type": "string",
      "minLength": 1,
      "description": "Optional stable model/token-space identifier for embedding artifacts. When omitted on every source, Antfly requires the effective producers to be semantically equivalent. To combine intentionally compatible but distinct producers, set the same identifier on every source. Explicit and implicit modes cannot be mixed; dimensions are always validated independently."
    },
    "chunk_size": {
      "x-go-type-skip-optional-pointer": true,
      "type": "integer",
      "minimum": 0,
      "description": "Chunk size for chunk enrichments."
    },
    "chunk_overlap": {
      "x-go-type-skip-optional-pointer": true,
      "type": "integer",
      "minimum": 0,
      "description": "Chunk overlap for chunk enrichments."
    },
    "chunker": {
      "$ref": "#/$defs/ChunkerConfig",
      "description": "Chunker configuration for chunk enrichments. Cannot be combined with chunker_json."
    },
    "chunker_json": {
      "x-go-type-skip-optional-pointer": true,
      "type": "string",
      "deprecated": true,
      "x-deprecated-reason": "Use the typed chunker field.",
      "description": "Legacy serialized chunker configuration for chunk enrichments. Cannot be combined with chunker."
    },
    "full_text_index": {
      "x-go-type-skip-optional-pointer": true,
      "type": "boolean",
      "default": false,
      "description": "When true on a chunk or asset enrichment, route generated text into the table's default full-text index."
    },
    "content_type": {
      "x-go-type-skip-optional-pointer": true,
      "type": "string",
      "description": "Produced asset content type for asset enrichments."
    },
    "producer": {
      "allOf": [
        {
          "$ref": "#/$defs/EnrichmentProducerConfig"
        }
      ],
      "writeOnly": true,
      "description": "Write-only producer configuration. Cannot be combined with producer_json or transcriber. Decision producers use type=decision and config={version, decider, questions}. A named-embedder similarity decider also requires embedders={name: frozen EmbedderConfig}; the dependency travels with the specification for deterministic replay instead of resolving mutable node configuration. Credentials use secret references or environment defaults. Outputs include answers, usage, resolved model, specification hash, version, and source fingerprint. Change version or specification to rebuild through the enrichment lifecycle."
    },
    "producer_json": {
      "x-go-type-skip-optional-pointer": true,
      "type": "string",
      "writeOnly": true,
      "deprecated": true,
      "x-deprecated-reason": "Use the producer field.",
      "description": "Write-only serialized producer configuration. For managed embedding enrichments Antfly stores a canonical semantic producer identity here; credentials and execution policy are excluded."
    },
    "neighbor_context": {
      "$ref": "#/$defs/EnrichmentNeighborContextConfig",
      "description": "Optional bounded sample of the document's graph neighbors appended to the producer input. Only valid on asset enrichments whose producer consumes rendered prompt text (generator, extractor, or decision); producers that treat the source as a media locator (copy, reader, transcriber, document_extraction) reject it. Only same-shard graph state is sampled; a graph index without local state for a document yields empty neighbors at runtime while the graph index reference itself is validated at admission."
    },
    "execution": {
      "$ref": "#/$defs/ExecutionPolicy",
      "description": "Non-semantic execution policy for this enrichment producer. This does not participate in generated artifact identity."
    },
    "transcriber": {
      "$ref": "#/$defs/TranscriberEnrichmentConfig",
      "description": "Typed shorthand for a transcription asset enrichment. Only valid with kind=asset and without producer_json; Antfly expands it into a document_extraction producer whose audio route transcribes each recording with this speech-to-text provider. The produced units carry the transcript text, provider confidence, and per-phrase time offsets, and chunk enrichments that consume them emit _start_time_ms/_end_time_ms on every chunk. With diarization: true the phrases also carry who spoke them, and a chunk that does not straddle a turn emits _speaker. content_type defaults to application/json."
    }
  },
  "$defs": {
    "AntflyChunkerConfig": {
      "description": "Configuration for the Antfly inference chunking provider.\n\nAntfly inference is Antfly's built-in ML service for local chunking.\n\n**Chunking Models:**\n- fixed: Fixed-size chunking by token count, with fixed time windows for audio\n  (aliases `fixed_bert`, `fixed_bpe`, and `fixed-bert-tokenizer`). This is the\n  only chunking model Antfly inference implements; other model names are\n  rejected as unsupported.\n\n**Deduplication:**\n- Within a single document write, chunk results are deduplicated when multiple\n  indexes share the same source text and chunker configuration, so the source\n  is chunked at most once per write.\n",
      "allOf": [
        {
          "$ref": "#/$defs/ChunkOptions"
        },
        {
          "type": "object",
          "properties": {
            "api_url": {
              "type": "string",
              "format": "uri",
              "description": "The URL of the Inference API endpoint (e.g., 'http://localhost:8080'). Can also be set via ANTFLY_INFERENCE_URL environment variable."
            },
            "model": {
              "type": "string",
              "description": "The chunking model to use. Defaults to 'fixed' for simple token-based chunking; other values select a model from models/chunkers/{name}/. Successful create responses include the effective model.",
              "default": "fixed"
            }
          }
        }
      ]
    },
    "AntflyGeneratorConfig": {
      "type": "object",
      "description": "Configuration for the Antfly inference generative AI provider.\n",
      "properties": {
        "provider": {
          "type": "string",
          "enum": [
            "antfly"
          ]
        },
        "model": {
          "type": "string",
          "description": "The name of the generator model."
        },
        "api_url": {
          "type": "string",
          "format": "uri",
          "description": "The URL of the Inference API endpoint. Can also be set via ANTFLY_INFERENCE_URL environment variable."
        },
        "temperature": {
          "type": "number",
          "format": "float",
          "description": "Controls randomness in generation (0.0-2.0).",
          "minimum": 0.0,
          "maximum": 2.0
        },
        "max_tokens": {
          "type": "integer",
          "description": "Maximum number of tokens to generate."
        },
        "top_p": {
          "type": "number",
          "format": "float",
          "description": "Nucleus sampling parameter.",
          "minimum": 0.0,
          "maximum": 1.0
        },
        "top_k": {
          "type": "integer",
          "description": "Top-k sampling parameter."
        },
        "timeout": {
          "type": "integer",
          "description": "HTTP response timeout in seconds for Inference API calls.",
          "minimum": 1
        }
      },
      "required": [
        "provider",
        "model"
      ]
    },
    "AntflySTTConfig": {
      "type": "object",
      "required": [
        "model"
      ],
      "description": "Configuration for Antfly inference STT (Whisper, Wav2Vec2, HuBERT) provider.\n\nUses the Antfly inference service for speech-to-text inference.\n\n**Supported Models:** openai/whisper-tiny, openai/whisper-base, facebook/wav2vec2-base\n\n**Supported Formats:** WAV (recommended), MP3, FLAC, M4A/AAC\n\n**Docs:** See inference documentation",
      "properties": {
        "api_url": {
          "type": "string",
          "format": "uri",
          "description": "Inference API URL. Falls back to ANTFLY_INFERENCE_URL environment variable."
        },
        "model": {
          "type": "string",
          "minLength": 1,
          "description": "Explicit Antfly transcriber model name (e.g., 'openai/whisper-tiny')."
        }
      }
    },
    "AnthropicGeneratorConfig": {
      "type": "object",
      "description": "Configuration for the Anthropic generative AI provider.\n",
      "properties": {
        "provider": {
          "type": "string",
          "enum": [
            "anthropic"
          ]
        },
        "workspace_id": {
          "type": "string",
          "description": "Anthropic workspace for keys not scoped to a workspace. Defaults to ANTHROPIC_WORKSPACE_ID."
        },
        "model": {
          "type": "string",
          "description": "The full model ID of the Anthropic model to use.",
          "default": "claude-sonnet-4-5-20250929"
        },
        "api_key": {
          "type": "string",
          "writeOnly": true,
          "description": "Anthropic API key or secret reference. Defaults to ANTHROPIC_API_KEY."
        },
        "url": {
          "type": "string",
          "format": "uri",
          "description": "Messages API base URL. Defaults to https://api.anthropic.com/v1."
        },
        "temperature": {
          "type": "number",
          "format": "float",
          "description": "Controls randomness in generation (0.0-1.0).",
          "minimum": 0.0,
          "maximum": 1.0
        },
        "max_tokens": {
          "type": "integer",
          "description": "Maximum number of tokens to generate in the response."
        },
        "top_p": {
          "type": "number",
          "format": "float",
          "description": "Nucleus sampling parameter (0.0-1.0).",
          "minimum": 0.0,
          "maximum": 1.0
        },
        "top_k": {
          "type": "integer",
          "description": "Top-k sampling parameter."
        }
      },
      "required": [
        "provider",
        "model"
      ]
    },
    "AppleGeneratorConfig": {
      "type": "object",
      "description": "On-device Apple Foundation Models generation. Requires a macOS Apple provider build, Apple Intelligence enabled, and its system model ready. Supports text conversations; tool calling and media attachments are not supported.",
      "properties": {
        "provider": {
          "type": "string",
          "enum": [
            "apple"
          ]
        },
        "max_tokens": {
          "type": "integer",
          "minimum": 1,
          "default": 256
        },
        "temperature": {
          "type": "number",
          "minimum": 0,
          "maximum": 2
        }
      }
    },
    "AppleSTTConfig": {
      "type": "object",
      "description": "On-device SpeechAnalyzer file transcription. Requires a macOS Apple provider build and supported Speech assets. Diarization is unsupported.",
      "properties": {
        "language_code": {
          "type": "string",
          "default": "en-US",
          "description": "Supported Speech locale; Apple transcription does not infer a language."
        },
        "timestamps": {
          "type": "boolean",
          "x-go-type-skip-optional-pointer": false,
          "default": true
        },
        "download_assets": {
          "type": "boolean",
          "default": false,
          "description": "Explicitly permit preparing and installing the selected Speech assets."
        }
      }
    },
    "AudioChunkOptions": {
      "type": "object",
      "description": "Options specific to audio chunking.",
      "properties": {
        "window_duration_ms": {
          "type": "integer",
          "minimum": 0,
          "maximum": 86400000,
          "description": "Window duration in milliseconds for fixed-window audio chunking (default: 30000)."
        },
        "overlap_duration_ms": {
          "type": "integer",
          "minimum": 0,
          "maximum": 86400000,
          "description": "Overlap duration in milliseconds between audio chunks (default: 0)."
        }
      }
    },
    "BedrockGeneratorConfig": {
      "type": "object",
      "description": "AWS Bedrock Converse generation, including image inputs and tools for\nmodels that support them. Uses the AWS default credential chain unless\nan explicit Bedrock API key is configured. Image capability is declared\nper target through inference_capabilities; model IDs are not allowlisted.\n",
      "properties": {
        "provider": {
          "type": "string",
          "enum": [
            "bedrock"
          ]
        },
        "url": {
          "type": "string",
          "description": "Optional Bedrock Runtime endpoint override. Defaults to the regional AWS endpoint."
        },
        "api_key": {
          "type": "string",
          "description": "Optional Bedrock bearer API key or secret reference. Omit to use signed AWS requests and the default credential chain."
        },
        "model": {
          "type": "string",
          "description": "The Bedrock model ID to use."
        },
        "region": {
          "type": "string",
          "description": "AWS region. Defaults to AWS_REGION, then AWS_DEFAULT_REGION, then us-east-1."
        },
        "temperature": {
          "type": "number",
          "format": "float",
          "description": "Controls randomness in generation (0.0-1.0).",
          "minimum": 0.0,
          "maximum": 1.0
        },
        "max_tokens": {
          "type": "integer",
          "description": "Maximum number of tokens to generate."
        },
        "top_p": {
          "type": "number",
          "format": "float",
          "description": "Nucleus sampling parameter.",
          "minimum": 0.0,
          "maximum": 1.0
        }
      },
      "required": [
        "provider",
        "model"
      ]
    },
    "ChatGPTGeneratorConfig": {
      "type": "object",
      "description": "Personal ChatGPT plan for interactive generation. Credentials stay in the local runtime.",
      "required": [
        "provider",
        "model",
        "connection_id"
      ],
      "properties": {
        "provider": {
          "type": "string",
          "enum": [
            "chatgpt"
          ]
        },
        "reasoning_effort": {
          "$ref": "#/$defs/OpenAIReasoningEffort"
        },
        "model": {
          "type": "string"
        },
        "connection_id": {
          "type": "string",
          "description": "Opaque personal registration owned by the authenticated caller."
        }
      }
    },
    "ChunkDiarizationConfig": {
      "type": "object",
      "description": "Source-wide online speaker clustering on a fixed 1.5 s grid with 3 s embedding windows. Identities are stable within a source and recipe. Silence and ambiguous voices remain unassigned. Analysis advances incrementally; admitted cached timelines and durable centroid/sample-position checkpoints are shared across matching outputs and continuations. Missing checkpoints reconstruct the required prefix deterministically.",
      "properties": {
        "model": {
          "type": "string",
          "description": "Speaker ONNX registry reference; defaults to 3D-Speaker CAM++. For kind=diarize the target model selects the speaker model."
        },
        "model_identity": {
          "type": "string",
          "pattern": "^[0-9a-f]{64}$",
          "description": "Optional SHA-256 pin of the speaker ONNX file, including when paired with a transcription model."
        },
        "similarity_threshold": {
          "type": "number",
          "format": "float",
          "minimum": 0,
          "maximum": 1,
          "default": 0.45
        },
        "ambiguity_margin": {
          "type": "number",
          "format": "float",
          "minimum": 0,
          "maximum": 1,
          "default": 0.03
        },
        "min_rms": {
          "type": "number",
          "format": "float",
          "minimum": 0,
          "maximum": 1,
          "default": 0.01
        },
        "max_speakers": {
          "type": "integer",
          "minimum": 1,
          "maximum": 64,
          "default": 64
        }
      }
    },
    "ChunkEmbeddingTarget": {
      "type": "object",
      "required": [
        "model"
      ],
      "properties": {
        "model": {
          "type": "string",
          "minLength": 1
        },
        "model_identity": {
          "type": "string",
          "minLength": 64,
          "maxLength": 64,
          "pattern": "^[0-9a-f]{64}$",
          "description": "Optional required embedding-weight/recipe identity; a changed model is rejected before delivery."
        },
        "task_type": {
          "type": "string",
          "default": "RETRIEVAL_DOCUMENT"
        },
        "dimensions": {
          "type": "integer",
          "minimum": 1
        },
        "execute": {
          "type": "boolean",
          "x-go-type-skip-optional-pointer": false,
          "default": true,
          "description": "False resolves model limits and plans descriptors without running embedding inference."
        },
        "allow_partial": {
          "type": "boolean",
          "default": false,
          "description": "Return explicit per-chunk failures and gaps instead of aborting on the first non-retryable embedding failure."
        }
      }
    },
    "ChunkOptions": {
      "type": "object",
      "description": "Per-request configuration for chunking. All fields are optional - zero/omitted values use chunker defaults.",
      "properties": {
        "max_chunks": {
          "type": "integer",
          "minimum": 0,
          "maximum": 4096,
          "description": "Maximum chunks in one response; zero requests the full result. Model-aware responses return continuation coverage when a page is full: follow next_continuation until source_complete. Embedded model-aware enrichment uses this as its page size and continues through the entire source. Fixed chunking without model-aware outputs retains its per-call truncation behavior."
        },
        "threshold": {
          "type": "number",
          "format": "float",
          "minimum": 0,
          "maximum": 1,
          "description": "Confidence threshold for model-based chunking (0.0-1.0)."
        },
        "text": {
          "$ref": "#/$defs/TextChunkOptions"
        },
        "audio": {
          "$ref": "#/$defs/AudioChunkOptions"
        },
        "embedding": {
          "$ref": "#/$defs/ChunkEmbeddingTarget"
        },
        "outputs": {
          "$ref": "#/$defs/ChunkOutputs"
        },
        "policy": {
          "$ref": "#/$defs/MediaChunkPolicy"
        }
      }
    },
    "ChunkOutputTarget": {
      "type": "object",
      "required": [
        "name",
        "kind",
        "model"
      ],
      "properties": {
        "name": {
          "type": "string",
          "minLength": 1,
          "maxLength": 128
        },
        "kind": {
          "type": "string",
          "enum": [
            "embedding",
            "transcribe",
            "classify",
            "diarize"
          ]
        },
        "model": {
          "type": "string",
          "minLength": 1,
          "maxLength": 4096
        },
        "execute": {
          "type": "boolean",
          "default": true,
          "description": "False retains this model as a planning boundary while skipping its inference; used for independent output repairs."
        },
        "skip_windows": {
          "type": "array",
          "maxItems": 4096,
          "description": "Execution-only repair mask for already accepted sub-windows; does not affect the model recipe or boundary plan. Skipped windows return receipts with status skipped.",
          "items": {
            "type": "object",
            "required": [
              "ordinal",
              "sub_index"
            ],
            "properties": {
              "ordinal": {
                "type": "integer",
                "format": "uint64",
                "minimum": 0
              },
              "sub_index": {
                "type": "integer",
                "format": "uint32",
                "minimum": 0
              }
            }
          }
        },
        "required": {
          "type": "boolean",
          "default": true
        },
        "model_identity": {
          "type": "string",
          "pattern": "^[0-9a-f]{64}$"
        },
        "task_type": {
          "type": "string",
          "default": "RETRIEVAL_DOCUMENT"
        },
        "dimensions": {
          "type": "integer",
          "minimum": 1
        },
        "language": {
          "type": "string"
        },
        "diarize": {
          "type": "boolean",
          "default": false,
          "description": "Convenience option on transcription outputs; enables source-wide diarization with the default speaker model."
        },
        "diarization": {
          "description": "Speaker settings for diarize or transcribe outputs; also accepted on the selected audio embedding boundary with audio_speaker_turns=true.",
          "$ref": "#/$defs/ChunkDiarizationConfig"
        },
        "top_k": {
          "type": "integer",
          "minimum": 1,
          "maximum": 128,
          "default": 5
        }
      }
    },
    "ChunkOutputs": {
      "type": "object",
      "required": [
        "boundary_model",
        "targets"
      ],
      "description": "Named independent media outputs. The named boundary model alone determines the shared plan; adding or changing another output preserves parent IDs. The compatibility embedding target is named embedding. Smaller output windows are identified by the parent ordinal and sub_index.",
      "properties": {
        "boundary_model": {
          "type": "string",
          "minLength": 1
        },
        "targets": {
          "type": "array",
          "minItems": 1,
          "maxItems": 16,
          "items": {
            "$ref": "#/$defs/ChunkOutputTarget"
          }
        }
      }
    },
    "ChunkerConfig": {
      "description": "A unified configuration for a chunking provider.",
      "allOf": [
        {
          "$ref": "#/$defs/AntflyChunkerConfig"
        },
        {
          "type": "object",
          "properties": {
            "provider": {
              "$ref": "#/$defs/ChunkerProvider"
            },
            "store_chunks": {
              "x-go-type-skip-optional-pointer": true,
              "type": "boolean",
              "description": "Controls whether chunk data is persisted to storage. When false (default), fixed chunks are generated in memory and only embeddings are stored. Model-aware streams always retain descriptor recipes; false omits their payloads. When true, chunk payloads are also stored. Vectors remain in the embedding artifact family.",
              "default": false
            },
            "full_text_index": {
              "x-go-type-skip-optional-pointer": true,
              "type": "object",
              "description": "Configuration for full-text indexing of chunks.\nWhen present (even if empty), chunk artifacts are persisted and indexed in Antfly's native full-text index, queryable and projectable via the document's `_chunks` field.\nWhen absent, chunks are generated only to drive vector embeddings and are not indexed for full-text search (unless `store_chunks` is also set).\n",
              "additionalProperties": true
            },
            "api_url": {
              "type": "string",
              "format": "uri",
              "description": "The URL of the Inference API endpoint. Can also be set via ANTFLY_INFERENCE_URL."
            },
            "model": {
              "type": "string",
              "description": "The chunking model to use. Defaults to 'fixed' for simple token-based chunking; other values select a model from models/chunkers/{name}/. Successful create responses include the effective model.",
              "default": "fixed"
            },
            "max_chunks": {
              "type": "integer",
              "minimum": 0,
              "maximum": 4096,
              "description": "Maximum chunks per invocation. Model-aware enrichment treats this as a page size and resumes until source completion; zero selects the worker page size. Fixed chunking retains its per-document cap, with zero meaning unlimited."
            },
            "threshold": {
              "type": "number",
              "format": "float",
              "minimum": 0,
              "maximum": 1
            },
            "text": {
              "$ref": "#/$defs/TextChunkOptions"
            },
            "audio": {
              "$ref": "#/$defs/AudioChunkOptions"
            },
            "embedding": {
              "$ref": "#/$defs/ChunkEmbeddingTarget"
            },
            "outputs": {
              "$ref": "#/$defs/ChunkOutputs"
            },
            "policy": {
              "$ref": "#/$defs/MediaChunkPolicy"
            }
          },
          "required": [
            "provider"
          ]
        }
      ]
    },
    "ChunkerProvider": {
      "type": "string",
      "enum": [
        "mock",
        "antfly"
      ],
      "description": "The chunking provider to use."
    },
    "CloudflareGeneratorConfig": {
      "type": "object",
      "description": "Cloudflare Workers AI generation. Uses CLOUDFLARE_API_TOKEN or an Antfly secret reference.",
      "properties": {
        "provider": {
          "type": "string",
          "enum": [
            "cloudflare"
          ]
        },
        "account_id": {
          "type": "string",
          "minLength": 1
        },
        "gateway_id": {
          "type": "string",
          "minLength": 1,
          "description": "Optional AI Gateway ID, sent through cf-aig-gateway-id."
        },
        "model": {
          "type": "string",
          "default": "@cf/meta/llama-3.3-70b-instruct-fp8-fast"
        },
        "api_key": {
          "type": "string",
          "writeOnly": true
        },
        "url": {
          "type": "string",
          "description": "Optional deployment-owned endpoint override."
        },
        "max_tokens": {
          "type": "integer",
          "minimum": 1
        },
        "temperature": {
          "type": "number",
          "format": "float",
          "minimum": 0,
          "maximum": 2
        },
        "top_p": {
          "type": "number",
          "format": "float",
          "minimum": 0,
          "maximum": 1
        },
        "frequency_penalty": {
          "type": "number",
          "format": "float",
          "minimum": -2,
          "maximum": 2
        },
        "presence_penalty": {
          "type": "number",
          "format": "float",
          "minimum": -2,
          "maximum": 2
        }
      },
      "required": [
        "provider",
        "account_id",
        "model"
      ]
    },
    "CloudflareSTTConfig": {
      "type": "object",
      "description": "Cloudflare Workers AI native speech adapter. Credentials use CLOUDFLARE_API_TOKEN.",
      "required": [
        "provider",
        "account_id",
        "model"
      ],
      "properties": {
        "provider": {
          "type": "string",
          "enum": [
            "cloudflare"
          ]
        },
        "account_id": {
          "type": "string"
        },
        "gateway_id": {
          "type": "string"
        },
        "model": {
          "type": "string",
          "default": "@cf/openai/whisper-large-v3-turbo"
        },
        "api_key": {
          "type": "string",
          "writeOnly": true
        },
        "base_url": {
          "type": "string",
          "description": "Optional deployment-owned full native endpoint override."
        },
        "language_code": {
          "type": "string"
        },
        "voice": {
          "type": "string"
        }
      }
    },
    "ConstrainedGenerationBatch": {
      "type": "object",
      "additionalProperties": false,
      "properties": {
        "max_items_per_call": {
          "type": "integer",
          "minimum": 1,
          "maximum": 256,
          "default": 8
        },
        "max_input_bytes": {
          "type": "integer",
          "minimum": 1,
          "maximum": 4194304,
          "default": 65536
        }
      }
    },
    "ConstrainedGeneratorProducerConfig": {
      "type": "object",
      "additionalProperties": false,
      "required": [
        "generator",
        "output_schema"
      ],
      "properties": {
        "generator": {
          "$ref": "#/$defs/GeneratorConfig"
        },
        "instructions": {
          "type": "string",
          "maxLength": 65536
        },
        "output_schema": {
          "type": "object",
          "additionalProperties": true,
          "description": "JSON schema enforced before artifact publication. Unsupported schema vocabulary fails admission. Invalid or missing batch items receive one validation retry; successful siblings are retained."
        },
        "output": {
          "$ref": "#/$defs/ConstrainedOutput"
        },
        "batch": {
          "$ref": "#/$defs/ConstrainedGenerationBatch"
        }
      }
    },
    "ConstrainedOutput": {
      "type": "string",
      "enum": [
        "auto",
        "tool_call",
        "json_schema",
        "grammar"
      ],
      "default": "auto",
      "description": "Provider and model admission selects a supported constrained mechanism. Auto prefers forced tools, then provider JSON schema, then native grammar. Free-text parsing is not a mode."
    },
    "ConstrainedProbabilities": {
      "type": "string",
      "enum": [
        "token_logprobs",
        "verdict_only"
      ],
      "default": "token_logprobs",
      "description": "Token probabilities require complete decoder evidence for the allowed values. Verdict-only is explicit opt-in; model-reported confidence is never used."
    },
    "DocumentExtractionConfig": {
      "type": "object",
      "description": "Configuration for the `document_extraction` producer. It downloads the source URL and persists derived units such as PDF pages, sections, images, and audio.\n\nNative text (PDF text layers, Office documents, HTML, email, plain text) is extracted with no mo

# --- truncated at 32 KB (125 KB total) ---
# Full source: https://artifacts.apis.io/antfly/json-schema/antfly-enrichment-config-schema.json

Work with this as data

Every JSON Schema here is available over the APIs.io API and to AI agents over MCP.

MCP server

One button, every client — Claude, Cursor, VS Code and the rest.

https://apis.io/mcp

Tools for schemas

4 MCP tools reach this
  • find_json_schemasBrowse and filter every JSON Schema in the catalog.
  • apis_io_searchSTART HERE — APIs, providers and tags for one query, each with its total.
  • resolveTurn a domain, URL or GitHub org into the provider it belongs to.
  • find_cohortsEvery scored population of providers in the catalog.
All 92 tools →

Call it yourself

curl for this page
This JSON Schema
curl "https://apis.io/api/v1/json-schemas/antfly-enrichment-config"
All schemas
curl "https://apis.io/api/v1/json-schemas?limit=25"

Calls without X-API-Key need no key. Calls with it need a paid plan — get a key.

Get an API key

Free tier, no form to fill in. Signing in shares your email address with us — we store it to create your key and to recognise you if you sign in with another provider. See our Privacy Policy and Terms.

A second provider on the same verified email joins the account you already have.