> ## Documentation Index > Fetch the complete documentation index at: https://docs.mathfi.ai/llms.txt > Use this file to discover all available pages before exploring further. # Feature refinement progress ## OpenAPI ````yaml api-reference/openapi.yaml GET /api/v1/feature-refinery/{featureRefinementKey}/progress openapi: 3.0.1 info: title: MathFi.ai API description: > The MathFi.ai REST API runs the two products end to end. **Feature Refinery** takes your labelled data and works out which columns actually carry the signal. It returns a reduced train and test pair, a per-column keep/remove decision, and the performance target the refined data reached. **Model Crucible** takes a dataset, trains every algorithm family against it at once, ranks the results on held-out data by fewest wrong decisions, and keeps the best three as versions of one model. You pick which one predictions run against. Everything is asynchronous: you create a thing, upload to a signed URL, start it, then poll until the state is terminal. Nothing streams and nothing blocks. contact: name: MathFi.ai url: https://mathfi.ai email: support@mathfi.ai license: name: MathFi.ai url: https://mathfi.ai version: 1.0.0 servers: - url: https://{tenant}-api.mathfi.ai description: >- Your tenant's API. Each customer has their own, so the host varies. Replace {tenant} with the name issued when your tenant was created. variables: tenant: default: your-tenant description: The tenant name issued to you. security: - BearerAuth: [] tags: - name: Authentication description: Exchange credentials for a bearer token - name: Feature Refinery description: Reduce a dataset to the columns that earn their place - name: Datasets description: >- Prepare labelled data for training, from uploads or from a finished refinement - name: Training description: Run the Crucible against a dataset and choose the champion model - name: Models description: Trained models and their versions - name: Predictions description: Score unlabelled data against a champion model paths: /api/v1/feature-refinery/{featureRefinementKey}/progress: get: tags: - Feature Refinery summary: Feature refinement progress operationId: getFeatureRefinementProgress parameters: - name: featureRefinementKey in: path required: true schema: type: string format: uuid responses: '200': description: Progress content: application/json: schema: $ref: '#/components/schemas/FeatureRefineryProgressOutput' '404': description: Unknown feature refinement key '500': description: Internal server error components: schemas: FeatureRefineryProgressOutput: description: >- Progress of a feature refinement — the current state, elapsed time, a recent activity trail, and the baseline vs current top performances, which together show how refinement is improving the model. type: object properties: featureRefinementKey: type: string format: uuid state: type: string description: >- The run's lifecycle state. It advances PENDING → SETUP_* → MEASURE_INITIAL_THRESHOLD* → BASELINE_RUN_* → REFINING_LOOP_* → CREATING_RESULT → REFINEMENT_COMPLETED. The terminal states are REFINEMENT_COMPLETED (success), ERROR and TIMEOUT (failures), and CANCELLED; once terminal, the state no longer changes and elapsedSeconds is frozen. Poll until the state is terminal. createdOn: type: string description: >- When the refinement was created, ISO-8601 UTC. Present from creation, so a run's age is available while it is still pending. elapsedSeconds: type: integer format: int64 description: >- Seconds the run has been processing — counted from when it started, not from creation. 0 while pending. progressCompleted: type: integer nullable: true description: >- Units of processing work completed so far, out of progressTotal. It increases by one as each unit finishes and never decreases; it is null (or 0) until the work has begun, and equals progressTotal once the run completes. Pair it with progressTotal for a determinate progress ratio. The unit is an internal measure with no standalone meaning and may change, so rely only on the ratio of the two and never on the raw numbers. progressTotal: type: integer nullable: true description: >- Total units of processing work for the run — the denominator for progressCompleted. Fixed for the life of the run once it is known; null while it is not yet known, in which case treat progress as indeterminate. cancelRequested: type: boolean nullable: true description: >- True once a cancel has been requested for this run, and it stays set — it is NOT cleared when the run reaches its terminal CANCELLED state, so a cancelled run keeps cancelRequested=true. It is therefore safe to rely on after a restart, rather than remembering the request locally. Treat it as meaningful only while the state is non-terminal; once terminal, state is authoritative. Omitted (null) when no cancel was ever requested. errorReason: type: string nullable: true description: >- Why the run stopped — present only when state is ERROR or TIMEOUT, null otherwise. Taken from the failure entry on the activity trail. recentActivities: type: array items: $ref: '#/components/schemas/ProgressActivity' baselineTopPerformances: description: >- Top model performances before refining, using all columns. Empty until the baseline is computed. type: array items: $ref: '#/components/schemas/PerformanceRow' currentTopPerformances: description: >- Top model performances on the feature set as it currently stands. Equals the baseline at the start and improves as the run progresses. type: array items: $ref: '#/components/schemas/PerformanceRow' ProgressActivity: description: >- One lifecycle entry on the public progress trail — the state and when it was reached. No free-text detail: the failure summary is on the progress errorReason field, and the full per-entry detail is kept on the ops-token view. type: object properties: timestamp: type: string description: ISO-8601 UTC instant (e.g. 2026-07-28T00:33:38.028Z). state: type: string PerformanceRow: description: >- One model's performance on the blind test set. Top-performance results are the best distinct results, ranked by fewest false decisions. type: object properties: falseDecisions: type: integer description: Number of wrong predictions on the test set (fewer is better). accuracy: type: number format: double nullable: true description: Null when the metric is undefined for this model (see recall). f1Score: type: number format: double nullable: true description: Null when the metric is undefined for this model (see recall). precision: type: number format: double nullable: true description: Null when the metric is undefined for this model (see recall). recall: type: number format: double nullable: true description: >- Null when the metric is undefined rather than zero — e.g. a test set slice with no positive cases leaves recall undefined, which is not the same as a model that recalled none of them. Treat an undefined metric as absent, never as 0. best: type: boolean description: >- True for the single best row in its table (the one with the fewest false decisions). securitySchemes: BearerAuth: type: http scheme: bearer bearerFormat: JWT ````