--- title: Fireworks AI API Integration on Truto slug: fireworks category: Artificial Intelligence canonical: "https://truto.one/integrations/detail/fireworks/" --- # Fireworks AI API Integration on Truto **Category:** Artificial Intelligence **Status:** Beta ## MCP-ready AI tools Truto exposes 206 tools for Fireworks AI that AI agents can call directly. - **list_all_fireworks_ai_accounts** — List Fireworks accounts ordered by name. Returns: name, displayName, createTime, accountType, email, state, status, suspendState, updateTime, notificationSettings, id, attributes. Supports an AIP-160 filter expression and a readMask to select returned fields. - **get_single_fireworks_ai_account_by_id** — Get a single Fireworks account by id. Returns: name, email, state, status, id. Use readMask to select which fields are returned. Required: id. - **fireworks_ai_accounts_test_eval** — Run a one-off test evaluation in Fireworks without creating an evaluation entry first. Returns the preview evaluation response including totalSamples and totalRuntimeMs summary statistics. Required: id, and a JSON request body defined by the upstream GatewayTestEvaluationBody schema. - **create_a_fireworks_ai_auth_refresh** — Refresh a session JWT in Fireworks before expiry with a fresh TTL. Returns a refreshed session token response. Required: a JSON request body containing the existing session token to refresh. - **fireworks_ai_credit_codes_redeem** — Redeem a credit code for a Fireworks account. Returns a confirmation response for the redemption. Consult the upstream Fireworks API reference for the request and response field contracts, as the field-level schema is not enumerated in the discovered source. - **create_a_fireworks_ai_validate_model_config** — Validate a model configuration in Fireworks. Returns an empty object on success. The request body carries the model config to validate; consult the Fireworks API reference for the specific fields of the validateModelConfig request payload. - **list_all_fireworks_ai_account_account_usage_filter_options** — List distinct usage filter options for a Fireworks account within a time range, covering serverless and dedicated deployment dimensions. Returns a filter_options map of dimension names to their distinct values. Required: account_id, startTime, endTime. - **list_all_fireworks_ai_account_audit_logs** — List user audit logs for a Fireworks account. Returns: auditLogs (array of audit log records), nextPageToken. Required: account_id. - **list_all_fireworks_ai_account_batch_inference_jobs** — List batch inference jobs for a Fireworks account. Returns a list of batch inference job objects, each including its name. Required: account_id. Supports AIP-160 filter grammar and custom ordering; defaults to ordering by created_time ascending. - **create_a_fireworks_ai_account_batch_inference_job** — Create a batch inference job in a Fireworks account. Returns the created batch inference job object including its name. Required: account_id. Optionally supply batchInferenceJobId as a query parameter to set the job ID; otherwise one is generated. - **delete_a_fireworks_ai_account_batch_inference_job_by_id** — Delete a batch inference job in Fireworks. Returns an empty object on success. Required: id, account_id. - **get_single_fireworks_ai_account_batch_inference_job_by_id** — Get a single batch inference job in Fireworks by id. Returns the batch inference job object including its name. Required: account_id, id. Use readMask to select specific fields. - **list_all_fireworks_ai_billing_summaries** — Get billing summary information for a Fireworks account over a given period. Returns the billing summary object with cost/usage fields aggregated daily for the requested range. Required: account_id, startTime, endTime. Note: only the date portion (YYYY-MM-DD) of each timestamp is used; endTime is exclusive (costs for the end date are not included). - **list_all_fireworks_ai_account_billing_usages** — List billing usage for a Fireworks account aggregated into daily buckets. Returns usage items, each carrying a group map of dimension values and cost data for the requested window. Required: account_id, startTime, endTime. The time window cannot exceed 31 days. - **fireworks_ai_account_checkpoints_promote** — Promote a checkpoint to a model in Fireworks. The trainer job ID in the request body resolves the GCS bucket. Returns a success response. Required: account_id, checkpoint_id, trainerJobId. - **list_all_fireworks_ai_account_clusters** — List Fireworks clusters for an account. Returns each cluster's name, displayName, state, createTime, updateTime, eksCluster, fakeCluster, and status. Supports AIP-160 filter grammar and custom orderBy. - **create_a_fireworks_ai_account_cluster** — Create a new cluster in a Fireworks account. Returns the created cluster object including its name, displayName, state, createTime, and updateTime. Required: account_id. - **update_a_fireworks_ai_account_cluster_by_id** — Update a Fireworks cluster. Returns the updated cluster object including its name, displayName, state, createTime, and updateTime. Required: account_id, cluster_id. The cluster.name field must be populated with the updated resource's name. - **delete_a_fireworks_ai_account_cluster_by_id** — Delete a Fireworks cluster by id. Returns an empty 204 response on success. Required: id (proxy cluster id). - **get_single_fireworks_ai_account_cluster_by_id** — Get a single Fireworks cluster by id. Returns: name, displayName, createTime, eksCluster, fakeCluster, state, status, updateTime. Required: id. - **fireworks_ai_account_clusters_get_connection_info** — Retrieve connection settings for a Fireworks cluster to be put in kubeconfig. Returns: endpoint, caData. Required: cluster_id. - **list_all_fireworks_ai_account_credit_redemptions** — List credit redemptions for a Fireworks account. Returns a paginated collection of credit redemption records (shape is content-type-specific — consult the upstream docs for the per-record field breakdown). Required: account_id. - **list_all_fireworks_ai_account_datasets** — List datasets in a Fireworks account, with optional filtering, ordering, and field masking. Returns: name, displayName, createTime, state, status, exampleCount, userUploaded, evaluationResult, transformed, splitted, evalProtocol, externalUrl, format, createdBy, updateTime, sourceJobName, estimatedTokenCount, averageTurnCount. Required: account_id. Max 200 per page (default 50). - **create_a_fireworks_ai_account_dataset** — Create a new dataset in a Fireworks account. Returns the created dataset object including its name, displayName, createTime, externalUrl, format, createdBy, and updateTime. Required: account_id. - **update_a_fireworks_ai_account_dataset_by_id** — Update a dataset's mutable properties in a Fireworks account. Returns the updated dataset object including its name, displayName, updateTime, externalUrl, format, createdBy, and sourceJobName. Required: account_id, id. - **delete_a_fireworks_ai_account_dataset_by_id** — Delete a dataset from a Fireworks account. Returns an empty 204 response on success. Required: account_id, id. - **get_single_fireworks_ai_account_dataset_by_id** — Get a Fireworks dataset by id. Returns the dataset object including name, displayName, createTime, updateTime, state, status, exampleCount, externalUrl, format, createdBy, estimatedTokenCount, and averageTurnCount. Required: id. - **fireworks_ai_account_datasets_get_download_endpoint** — Get a signed download URL for a Fireworks dataset. Returns a url field pointing to the downloadable dataset archive. Required: dataset_id. Optional: download_lineage to download the entire lineage chain (filenames prefixed with dataset IDs). - **fireworks_ai_account_datasets_get_upload_endpoint** — Get a signed upload URL for uploading files to a Fireworks dataset. Use the returned url to PUT the dataset file (for files larger than 150MB; smaller files can use the single-request upload endpoint). Returns a url field. Required: dataset_id. - **fireworks_ai_account_datasets_split_dataset** — Split a Fireworks dataset into subsets (e.g. train/validation/test). Returns a dataset split result object. Required: account_id, dataset_id. - **fireworks_ai_account_datasets_validate_upload** — Validate a Fireworks dataset upload. Returns an empty object on success. Required: account_id, dataset_id. - **fireworks_ai_account_datasets_upload** — Upload a dataset file to a Fireworks dataset in a single multipart request. Handles files up to 150MB — for larger files use the Get Dataset Upload Endpoint to obtain a presigned URL. Returns a FileUploadResponse object with content-type-specific fields (the schema is referenced but not enumerated in the public docs). Required: account_id, dataset_id, file. - **list_all_fireworks_ai_account_deployed_models** — List LoRA models deployed to a Fireworks account. Returns: name, displayName, description, createTime, model, deployment, state, serverless, status, public, updateTime. Supports filter (AIP-160 grammar), orderBy, and readMask. - **create_a_fireworks_ai_account_deployed_model** — Load (deploy) a LoRA model to a Fireworks account. Returns the created deployed model including name, displayName, model, deployment, state, status, serverless, public, and default. Required: account_id. Optional query: replaceMergedAddon for hot reload deployments. - **update_a_fireworks_ai_account_deployed_model_by_id** — Update a LoRA deployed model in a Fireworks account. Returns the updated deployed model including name, displayName, model, deployment, state, status, serverless, public, and default. Required: id. The request body's `name` must be populated with the updated resource's name. - **delete_a_fireworks_ai_account_deployed_model_by_id** — Unload (delete) a LoRA deployed model from a Fireworks account. Returns an empty response on success. Required: id. - **get_single_fireworks_ai_account_deployed_model_by_id** — Get a deployed LoRA model by id. Returns: name, displayName, description, createTime, model, deployment, state, serverless, status, public, updateTime. Required: account_id, id. - **list_all_fireworks_ai_account_deployment_shapes** — List deployment shapes for a Fireworks account, optionally filtered by target model compatibility and ordered by create_time. Returns an array of deployment shape records including name, displayName, description, baseModel, acceleratorCount, acceleratorType, precision, createTime, and updateTime. Max 200 per page. - **create_a_fireworks_ai_account_deployment_shape** — Create a deployment shape in a Fireworks account. Returns the created deployment shape including name, displayName, description, baseModel, acceleratorCount, acceleratorType, precision, createTime, and updateTime. Required: a JSON request body with the shape properties (e.g. baseModel). - **update_a_fireworks_ai_account_deployment_shape_by_id** — Update a deployment shape in Fireworks. Returns the updated deployment shape including name, displayName, description, baseModel, acceleratorType, acceleratorCount, precision, and createTime. Required: id, baseModel. - **delete_a_fireworks_ai_account_deployment_shape_by_id** — Delete a deployment shape in Fireworks by id. Returns an empty 204 response on success. Required: id, account_id. - **get_single_fireworks_ai_account_deployment_shape_by_id** — Get a deployment shape in Fireworks by id. Returns the deployment shape object including its displayName, baseModel, acceleratorType, acceleratorCount, createTime, and updateTime. Required: account_id, id. - **list_all_fireworks_ai_versions** — List deployment shape versions for a deployment shape in Fireworks. Returns: name, createTime, snapshot, validated, public, latestValidated. Required: account_id, deployment_shape_id. - **get_single_fireworks_ai_version_by_id** — Get a single deployment shape version in Fireworks. Returns the version object with content-type-specific fields. Required: account_id, deployment_shape_id, id. - **update_a_fireworks_ai_version_by_id** — Update a deployment shape version in Fireworks. Returns: name, createTime, snapshot, validated, public, latestValidated. Required: account_id, deployment_shape_id, id. - **list_all_fireworks_ai_account_deployments** — List deployments in a Fireworks account. Returns: name, displayName, description, createTime, expireTime, purgeTime, deleteTime, state, status, annotations, minReplicaCount, maxReplicaCount, maxWithRevocableReplicaCount, desiredReplicaCount, replicaCount, autoscalingPolicy, baseModel, acceleratorCount, acceleratorType, precision, cluster, enableAddons, draftTokenCount, draftModel,… - **create_a_fireworks_ai_account_deployment** — Create a deployment in a Fireworks account. Returns the created deployment object including name, baseModel, state, createTime, minReplicaCount, maxReplicaCount, status. Required: account_id. - **update_a_fireworks_ai_account_deployment_by_id** — Update a Fireworks deployment. Returns the updated deployment object including name, baseModel, state, updateTime, minReplicaCount, maxReplicaCount, status, displayName, region. Required: account_id, id. Body must include baseModel. - **delete_a_fireworks_ai_account_deployment_by_id** — Delete a Fireworks deployment by id. Supports hard deletion and ignoring in-use checks via query flags. Required: id. Returns an empty 204 response on success. - **get_single_fireworks_ai_account_deployment_by_id** — Get a single Fireworks deployment by id. Returns the deployment object including its id and attributes (deployment-specific fields defined upstream in the gatewayDeployment schema). Required: account_id, id. - **fireworks_ai_account_deployments_get_metrics** — Get metrics for a Fireworks deployment (deprecated). Returns a metrics response object including its id and attributes (metrics-specific fields defined upstream in the gatewayGetDeploymentMetricsResponse schema, controlled by readMask). Required: account_id, deployment_id. - **fireworks_ai_account_deployments_scale** — Scale a Fireworks deployment to a specific number of replicas or to zero. Returns an empty response on success. Required: account_id, deployment_id. - **fireworks_ai_account_deployments_undelete** — Undelete a previously deleted Fireworks deployment, restoring it to an active state. Returns the restored deployment object including its name, displayName, baseModel, state, status, minReplicaCount, maxReplicaCount, and placement. Required: account_id, deployment_id. - **list_all_fireworks_ai_deployment_ledgers** — Get the ledger for a Fireworks deployment (used for hot-load operations). Returns: ledger. Required: account_id, deployment_id. - **fireworks_ai_deployment_ledgers_bulk_delete** — Reset the ledger for hot load on a Fireworks deployment. Returns an empty object on success. Required: account_id, deployment_id. - **list_all_fireworks_ai_deployment_shards** — List deployment shards for a Fireworks deployment. Returns: next_page_token, shards (array of shard records including create_time and additional fields selectable via readMask). Required: account_id, deployment_id. - **list_all_fireworks_ai_deployment_termination_messages** — Get the termination message for a Fireworks deployment. Returns a response whose field-level shape is defined by the upstream gatewayGetTerminationMessageResponse schema and is not enumerated in the crawled Fireworks API reference; consult the upstream docs for the property breakdown. Required: account_id, deployment_id. - **list_all_fireworks_ai_account_developer_passes** — List developer passes for a Fireworks account. Returns: name, autoRenew, endTime, createTime, updateTime, state, lastRenewTime. Supports filter (e.g. "state=ACTIVE") and orderBy expressions. - **create_a_fireworks_ai_account_developer_pass** — Create a developer pass for a Fireworks account. Returns: name, state. - **update_a_fireworks_ai_account_developer_pass_by_id** — Update a developer pass. Returns: name, autoRenew, endTime, createTime, updateTime, state, lastRenewTime. Required: id. - **get_single_fireworks_ai_account_developer_pass_by_id** — Get a single developer pass by id. Returns: name, autoRenew, endTime, createTime, updateTime, state, lastRenewTime. Required: id. - **list_all_fireworks_ai_account_dpo_jobs** — List DPO jobs for a Fireworks account. Returns an array of DPO job objects including name, dpoJobId, and content-type-specific attributes. Supports AIP-160 filter syntax, custom ordering, and read masks. Required: account_id. - **create_a_fireworks_ai_account_dpo_job** — Create a new DPO job under a Fireworks account. Returns the created DPO job object including name, dpoJobId, and content-type-specific attributes. Optional dpoJobId query parameter assigns a specific ID; otherwise one is generated. Required: account_id. - **delete_a_fireworks_ai_account_dpo_job_by_id** — Delete a DPO job in Fireworks. Returns an empty JSON object on success. Required: account_id, id. - **get_single_fireworks_ai_account_dpo_job_by_id** — Get a DPO job by id in a Fireworks account. Returns the DPO job object including its id and content-type-specific attributes defined by the upstream gatewayDpoJob schema. Required: account_id, id. - **fireworks_ai_account_dpo_jobs_cancel** — Cancel a Fireworks DPO job. Returns an empty object on success. Required: dpo_job_id. - **fireworks_ai_account_dpo_jobs_get_metrics_file_endpoint** — Get the metrics file endpoint for a Fireworks DPO job. Returns an empty object on success. Required: dpo_job_id. - **fireworks_ai_account_dpo_jobs_resume** — Resume a paused DPO (Direct Preference Optimization) job in Fireworks. Returns the DPO job object including its resource name, dataset, state, and training configuration. Required: account_id, dpo_job_id. - **list_all_fireworks_ai_account_evaluation_jobs** — List evaluation jobs for a Fireworks account. Returns a paginated collection of evaluation job objects, each exposing content-type-specific fields (consult the upstream API reference for the full field-level breakdown). Required: account_id. - **create_a_fireworks_ai_account_evaluation_job** — Create a new evaluation job in Fireworks. Returns the created evaluation job object with content-type-specific fields (consult the upstream API reference for the full field-level breakdown). Required: account_id. - **delete_a_fireworks_ai_account_evaluation_job_by_id** — Delete an evaluation job in Fireworks. Required: account_id, id. Returns an empty 200 response on success. - **get_single_fireworks_ai_account_evaluation_job_by_id** — Get a Fireworks evaluation job by id. Returns the evaluation job object including its name, state, evaluator, inputDataset, outputDataset, createTime, and updateTime. Required: id. - **fireworks_ai_account_evaluation_jobs_get_execution_log_endpoint** — Get the execution log stream endpoint and tracing IDs for a Fireworks evaluation job. Returns a response object with its id and additional fields (log stream URL, tracing identifiers) defined in the upstream Fireworks schema. Required: evaluation_job_id. - **list_all_fireworks_ai_account_evaluations** — List evaluations for a Fireworks account. Returns evaluation objects with id and evaluation-specific fields. Supports server-side filtering, ordering, and a readMask to select returned fields. Required: account_id. - **create_a_fireworks_ai_account_evaluation** — Create an evaluation in a Fireworks account. Returns the created evaluation object including its id and evaluation-specific fields. Required: account_id. - **delete_a_fireworks_ai_account_evaluation_by_id** — Delete an evaluation in Fireworks AI. Returns an empty response on success. Required: account_id, id. - **fireworks_ai_account_evaluations_validate_assertions** — Validate evaluation assertions in Fireworks. Returns a validation response object whose fields are defined by the upstream GatewayValidateAssertionsResponse schema. Required: a JSON request body conforming to the GatewayValidateAssertionsBody schema. - **get_single_fireworks_ai_account_evaluation_by_id** — Get a single evaluation by id in Fireworks. Returns the evaluation object whose fields are defined by the upstream gatewayEvaluation schema. Required: id. Optional: readMask to select specific fields (defaults to all fields when empty or "*"). - **fireworks_ai_account_evaluations_preview** — Preview an evaluation with sample data in Fireworks. Returns a preview response object whose fields are defined by the upstream gatewayPreviewEvaluationResponse schema. Required: evaluation_id, a JSON request body conforming to the GatewayPreviewEvaluationBody schema. - **list_all_fireworks_ai_account_evaluators** — List evaluators for a Fireworks account with pagination support. Returns a collection of evaluator objects including their name, displayName, description, state, status, createTime, updateTime, criteria, requirements, entryPoint, source, and defaultDataset. Required: account_id. - **create_a_fireworks_ai_account_evaluator** — Create a new evaluator in a Fireworks account. Deprecated — use CreateEvaluatorV2 instead. Returns the created evaluator object including name, displayName, description, state, status, createTime, updateTime, criteria, requirements, entryPoint, source, and defaultDataset. Required: account_id and a JSON body describing the evaluator. - **update_a_fireworks_ai_account_evaluator_by_id** — Update evaluator metadata (displayName, description, defaultDataset) in Fireworks. Changing requirements or entryPoint triggers a rebuild; to upload new source code, set prepareCodeUpload: true. Returns the updated evaluator object including name, displayName, description, state, status, createTime, updateTime, criteria, requirements, entryPoint, source, defaultDataset. Required: account_id,… - **delete_a_fireworks_ai_account_evaluator_by_id** — Delete an evaluator and its associated versions and build artifacts from a Fireworks account. Returns an empty 204 response on success. Required: account_id, id. - **fireworks_ai_account_evaluators_preview_evaluator** — Preview an evaluator configuration before creation. Returns a preview result. Required: account_id. - **get_single_fireworks_ai_account_evaluator_by_id** — Get a Fireworks evaluator by id. Returns the evaluator object including its displayName, description, state (BUILDING / ACTIVE / BUILD_FAILED), status, requirements, entryPoint, criteria, commitHash, source, defaultDataset, createTime, createdBy, and updateTime. Use this to poll build progress after creation. Required: id. - **fireworks_ai_account_evaluators_get_build_log_endpoint** — Get a signed URL to download a Fireworks evaluator's build logs. Useful for debugging the BUILD_FAILED state. Returns the response containing the signed URL. Required: evaluator_id. - **fireworks_ai_account_evaluators_get_source_code_signed_url** — Get a signed URL to download a Fireworks evaluator's source code archive. Useful for debugging or reviewing the uploaded code. Returns filenameToSignedUrls mapping each source file to its signed download URL. Required: evaluator_id. - **fireworks_ai_account_evaluators_get_upload_endpoint** — Get signed URLs for uploading a Fireworks evaluator's source code (step 3 of the Create Evaluator workflow). After receiving the signed URL, PUT the .tar.gz archive with Content-Type: application/octet-stream. Returns filenameToSignedUrls mapping each source file to its signed upload URL. Required: evaluator_id, filenameToSize. - **fireworks_ai_account_evaluators_validate_upload** — Trigger server-side validation of an uploaded evaluator source archive (step 5 of the Create Evaluator workflow). The server extracts and processes the archive, then builds the evaluator environment; poll Get Evaluator to monitor progress. Returns an empty object on success. Required: evaluator_id. - **list_all_fireworks_ai_evaluator_versions** — List evaluator revisions (versions) for a Fireworks evaluator, ordered reverse-chronologically by default per AIP-162. Returns: id, attributes. Required: account_id, evaluator_id. - **delete_a_fireworks_ai_evaluator_version_by_id** — Delete an evaluator revision in Fireworks. If the supplied name is an alias rather than a concrete revision, the alias is deleted instead (per AIP-162). Returns an empty 204 response on success. Required: id, evaluator_id. - **get_single_fireworks_ai_evaluator_version_by_id** — Get a Fireworks evaluator version (revision) by account id. Returns the evaluator version object including its id and version-specific attributes. Required: account_id, evaluator_id, id. - **fireworks_ai_evaluator_versions_rollback** — Rollback a Fireworks evaluator to a specific revision. Returns the updated evaluator version object including its id and version-specific attributes. Required: account_id, evaluator_id, version_id. - **fireworks_ai_evaluator_versions_alias** — Assign or update a user-specified alias that points at a specific evaluator revision in Fireworks. Returns: name, snapshot. Required: account_id, evaluator_id, version_id. - **create_a_fireworks_ai_account_evaluators_v_2** — Create a custom evaluator for scoring model outputs in Fireworks. Evaluators use the Eval Protocol to define test cases, run model inference, and score responses for evaluation jobs and Reinforcement Fine-Tuning. Returns the created evaluator object with schema-specific fields. Required: account_id. - **list_all_fireworks_ai_account_generic_deployment_types** — List the generic deployment types under a Fireworks account. Returns items including createTime, updateTime, and imageTag (the helm chart image tag). Required: account_id. Max 200 per page; default order is by create_time. - **create_a_fireworks_ai_account_generic_deployment_type** — Create a generic deployment type in Fireworks. Returns the created generic deployment type version object including id, name, createTime, updateTime, and imageTag. Required: account_id, genericDeploymentTypeId. - **update_a_fireworks_ai_account_generic_deployment_type_by_id** — Update a generic deployment type in Fireworks. Returns the updated version including id, name, createTime, updateTime, and imageTag. Required: account_id, id. - **delete_a_fireworks_ai_account_generic_deployment_type_by_id** — Delete a generic deployment type in Fireworks. Returns an empty object on success. Required: account_id, id. - **get_single_fireworks_ai_account_generic_deployment_type_by_id** — List all versions of a generic deployment type in a Fireworks account. Returns a paginated list of generic deployment type versions including createTime, updateTime, and imageTag. Required: id (the generic deployment type id). - **list_all_fireworks_ai_account_generic_deployments** — List all generic deployments under a Fireworks account. Returns: name, releaseValues, placement, createTime, updateTime, genericDeploymentType, state. Supports AIP-160 filter grammar, custom orderBy, and readMask field selection. Required: account_id. - **create_a_fireworks_ai_account_generic_deployment** — Create a generic deployment in a Fireworks account. Returns the created deployment including genericDeploymentType, releaseValues, placement, createTime, updateTime, and state. Required: account_id, plus body fields releaseValues, placement, genericDeploymentType. Optional query: genericDeploymentId (random ID generated if omitted), validateOnly (preview without creating). - **update_a_fireworks_ai_account_generic_deployment_by_id** — Update a generic deployment in a Fireworks account. Returns the updated deployment including genericDeploymentType, releaseValues, placement, state, createTime, updateTime. Required: account_id, id, plus body fields releaseValues, placement, genericDeploymentType. - **delete_a_fireworks_ai_account_generic_deployment_by_id** — Delete a generic deployment from a Fireworks account. Required: id (proxy record id). Optional query: hard (perform hard deletion), ignoreChecks (force deletion of a deployment currently in use). - **get_single_fireworks_ai_account_generic_deployment_by_id** — Get a generic deployment in Fireworks by id. Returns: name, releaseValues, placement, createTime, updateTime, genericDeploymentType, state. Required: id. - **list_all_fireworks_ai_account_identity_providers** — List identity providers configured for a Fireworks account. Returns: name, displayName, createTime, updateTime, samlConfig, oidcConfig, tenantDomains, state, status, domainUrl, issuerUrl, clientId, enableJitUserProvisioning, jitDefaultRole, enforceSso, enableIdpInitiatedSso. Supports optional filter, orderBy, and readMask query parameters. Required: account_id. - **create_a_fireworks_ai_account_identity_provider** — Create a new identity provider for a Fireworks account. Returns the created identity provider including displayName, tenantDomains, samlConfig, oidcConfig, domainUrl, issuerUrl, clientId, enableJitUserProvisioning, jitDefaultRole, enforceSso, enableIdpInitiatedSso, createTime, and updateTime. Required: account_id. - **update_a_fireworks_ai_account_identity_provider_by_id** — Update an existing identity provider for a Fireworks account. Returns the updated identity provider including displayName, tenantDomains, samlConfig, oidcConfig, domainUrl, issuerUrl, clientId, enableJitUserProvisioning, jitDefaultRole, enforceSso, enableIdpInitiatedSso, createTime, and updateTime. Required: id. - **delete_a_fireworks_ai_account_identity_provider_by_id** — Delete an identity provider from a Fireworks account. Returns an empty 204 response on success. Required: id. - **get_single_fireworks_ai_account_identity_provider_by_id** — Get a single identity provider for a Fireworks account by id. Returns the identity provider object including its display_name, tenant_domains, saml_config, oidc_config, state, status, domain_url, issuer_url, client_id, enable_jit_user_provisioning, jit_default_role, enforce_sso, enable_idp_initiated_sso, create_time, and update_time. Required: account_id, id. - **list_all_fireworks_ai_account_models** — List models in a Fireworks account. Returns an array of model objects including name, displayName, state, kind, createTime, and updateTime. Required: account_id. Max 200 per page (default 50). - **create_a_fireworks_ai_account_model** — Create a new model in a Fireworks account. Returns the created model object including name, displayName, state, kind, and createTime. Required: account_id and a JSON body describing the model (model_id, displayName, kind, and source details such as huggingFaceUrl or githubUrl). - **update_a_fireworks_ai_account_model_by_id** — Update an existing model in a Fireworks account. Returns the updated model object including name, displayName, state, kind, and updateTime. Required: id (the model resource name, e.g. accounts/{account_id}/models/{model_id}). - **delete_a_fireworks_ai_account_model_by_id** — Delete a model in a Fireworks account by id. Returns an empty 200 response on success (no body content). Required: id (the model resource name, e.g. accounts/{account_id}/models/{model_id}). - **get_single_fireworks_ai_account_model_by_id** — Get a Fireworks model by id. Returns the full model object including displayName, description, state, kind, createTime, updateTime, contextLength, public, supportsImageInput, supportsTools, and other model metadata. Required: id. - **fireworks_ai_account_models_get_download_endpoint** — Get signed download URLs for a Fireworks model's files. Returns an object containing filenameToSignedUrls mapping file names to their signed download URLs. Required: model_id. - **fireworks_ai_account_models_get_upload_endpoint** — Get signed upload URLs for a Fireworks model's files. Returns an object containing filenameToSignedUrls and filenameToUnsignedUris mappings for uploading model files. Required: model_id. - **fireworks_ai_account_models_import** — Transfer a model from S3 to GCP storage in Fireworks. Returns a long-running operation object for tracking the async transfer (operation fields are not enumerated in the source). Required: account_id, model_id. - **fireworks_ai_account_models_prepare** — Prepare a Fireworks model for different precisions. Returns an empty 200 response on success. Required: model_id. - **fireworks_ai_account_models_validate_upload** — Validate a model upload in Fireworks. Returns the validation result object. Required: account_id, model_id. - **create_a_fireworks_ai_model_version** — Create a model version under a Fireworks account and model. Returns the created model version object. Required: account_id, model_id. - **fireworks_ai_model_versions_count** — Get the number of model versions for a Fireworks model. Returns a count response object. Required: account_id, model_id. - **fireworks_ai_model_versions_deploy** — Deploy a specific model version in Fireworks. Returns a long-running operation object. Required: account_id, model_id, version_id. - **list_all_fireworks_ai_account_policy_settings** — Get the singleton PolicySettings for a Fireworks account. Returns the policy settings object including its name. Required: account_id. - **fireworks_ai_account_policy_settings_bulk_update** — Update the per-account PolicySettings singleton in Fireworks. Returns the updated policy settings object including its name, rules, and updateTime. Required: account_id. - **list_all_fireworks_ai_account_pricing_plans** — List pricing plans for a Fireworks account. Returns: name, billingType, tokenConfig, acceleratorHourConfig, startTime, endTime, createTime, updateTime, state, id, attributes. Supports filtering (e.g. state=READY) and ordering (e.g. create_time desc). - **get_single_fireworks_ai_account_pricing_plan_by_id** — Get a single Fireworks pricing plan by id. Returns: name, billingType, tokenConfig, acceleratorHourConfig, startTime, endTime, createTime, updateTime, state, id, attributes. Required: id. - **list_all_fireworks_ai_account_quotas** — List all quotas for a Fireworks account. Returns: quotas. - **update_a_fireworks_ai_account_quota_by_id** — Update a quota. Returns: name, value, maxValue, usage, updateTime. Required: id. - **get_single_fireworks_ai_account_quota_by_id** — Get a single quota by id. Returns: name, value, maxValue, usage, updateTime. Required: id. - **list_all_fireworks_ai_account_reinforcement_fine_tuning_jobs** — List reinforcement fine-tuning jobs for a Fireworks account. Returns: name, displayName, createTime, completedTime, dataset, evaluationDataset, evalAutoCarveout, state, status, createdBy, trainingConfig, evaluator, wandbConfig, awsS3Config, azureBlobStorageConfig, outputStats, jobProgress, inferenceParameters, chunkSize, outputMetrics, maxInferenceReplicaCount, nodeCount, lossConfig,… - **create_a_fireworks_ai_account_reinforcement_fine_tuning_job** — Create a reinforcement fine-tuning job in Fireworks. Returns the created job including displayName, state, status, dataset, evaluator, createTime, and createdBy. Required in body: dataset, evaluator. Optional query: reinforcementFineTuningJobId to supply a custom job UUID (otherwise a random UUID is generated). - **update_account_reinforcement_fine_tuning_job_by_id** — Update a reinforcement fine-tuning job in Fireworks. Returns the updated job including displayName, state, status, dataset, evaluator, createTime, and trainingConfig. Required: id (proxy record id), dataset, evaluator. - **delete_account_reinforcement_fine_tuning_job_by_id** — Delete a reinforcement fine-tuning job in Fireworks. Returns an empty 204 response on success. Required: id (proxy record id). - **get_account_reinforcement_fine_tuning_job_by_id** — Get a reinforcement fine-tuning job by id in Fireworks. Returns the job object including name, displayName, createTime, dataset, state, status, evaluator, and purpose. Required: account_id, id. - **fireworks_ai_account_reinforcement_fine_tuning_jobs_cancel** — Cancel a reinforcement fine-tuning job in Fireworks. Returns an empty object on success. Required: reinforcement_fine_tuning_job_id. - **fireworks_ai_account_reinforcement_fine_tuning_jobs_debug** — Debug a reinforcement fine-tuning job in Fireworks. Returns a debug response object. Required: reinforcement_fine_tuning_job_id. - **account_reinforcement_fine_tuning_jobs_get_metrics_file_endpoint** — Get the metrics file endpoint for a reinforcement fine-tuning job in Fireworks. Returns a metrics file response object. Required: reinforcement_fine_tuning_job_id. - **fireworks_ai_account_reinforcement_fine_tuning_jobs_resume** — Resume a reinforcement fine-tuning job in Fireworks. Returns the resumed job object including state, status, dataset, evaluator, trainingConfig, and jobProgress. Required: reinforcement_fine_tuning_job_id. - **list_all_fireworks_ai_account_rlor_trainer_jobs** — List reinforcement fine-tuning (RLOR trainer) steps for a Fireworks account. Returns a paginated collection of RLOR trainer job objects (each exposing at least a name identifier; remaining fields are content-type-specific). Required: account_id. Max 200 per page (default 50). - **create_a_fireworks_ai_account_rlor_trainer_job** — Create a reinforcement fine-tuning (RLOR trainer) step in a Fireworks account. Returns the created RLOR trainer job object including its name and content-type-specific attributes (training config, datasets, loss config, progress, etc.). Required: account_id. Optional query parameters: rlorTrainerJobId (a random UUID is generated if omitted), trainingShape (validated training-shape selector for… - **delete_a_fireworks_ai_account_rlor_trainer_job_by_id** — Delete a reinforcement fine-tuning (RLOR trainer) job. Returns an empty 204 response on success. Required: account_id, rlor_trainer_job_id. - **get_single_fireworks_ai_account_rlor_trainer_job_by_id** — Get a Reinforcement Fine-tuning Step (RLOR Trainer Job) by id in Fireworks. Returns the RLOR trainer job object including its name, displayName, state, dataset, and training configuration. Required: id. - **fireworks_ai_account_rlor_trainer_jobs_cancel** — Cancel an RLOR trainer job in Fireworks. Returns an empty object on success. Required: account_id, rlor_trainer_job_id. - **fireworks_ai_account_rlor_trainer_jobs_execute_train_step** — Execute one training step for a keep-alive Reinforcement Fine-tuning (RLOR) job in Fireworks. Returns an empty object on success. Required: account_id, rlor_trainer_job_id. - **fireworks_ai_account_rlor_trainer_jobs_resume** — Resume a paused reinforcement fine-tuning step (RLOR trainer job) in Fireworks. Returns the resumed rlor trainer job object including its id, name, and content-type-specific attributes. Required: rlor_trainer_job_id. - **list_all_fireworks_ai_checkpoints** — List checkpoints for an RLOR Trainer Job in Fireworks. Returns an array of checkpoint objects, each containing an id and content-type-specific attributes. Required: account_id, rlor_trainer_job_id. Max 200 per page. - **list_all_fireworks_ai_public_logs** — Get a signed URL for the public trainer logs of an RLOR Trainer Job in Fireworks. Returns: url. Required: account_id, rlor_trainer_job_id. - **list_all_fireworks_ai_account_routers** — List routers for a Fireworks account. Returns: name, displayName, createTime, createdBy, state, status, deployments, model, weightedRandom, evenLoad, aliases, autoGenerated, public. Supports AIP-160 filter syntax, custom ordering, and field read masks. - **create_a_fireworks_ai_account_router** — Create a router for a Fireworks account. Returns the created router including name, displayName, state, status, deployments, model, and public. Optional: routerId to specify the router's id. - **update_a_fireworks_ai_account_router_by_id** — Update a router's properties. Returns the updated router including name, displayName, state, status, deployments, model, and public. Required: id, account_id. - **delete_a_fireworks_ai_account_router_by_id** — Delete a router by id. Returns an empty object on success. Required: id. - **get_single_fireworks_ai_account_router_by_id** — Get a Fireworks router by id within an account. Returns the router object including its name, displayName, state, status, deployments, model, aliases, autoGenerated, public, createTime, and createdBy. Required: id. Use readMask to restrict the fields returned; if empty or "*", all fields are returned. - **list_all_fireworks_ai_account_secrets** — List all secrets for a Fireworks account. Returns: name, keyName, value. The secret value is not included in responses for security reasons. Required: account_id. - **create_a_fireworks_ai_account_secret** — Create a new secret for a Fireworks account. Returns: name, keyName, value. The secret value is accepted on input but never returned in responses for security reasons. Required: account_id, keyName. - **update_a_fireworks_ai_account_secret_by_id** — Update an existing secret for a Fireworks account. Returns: name, keyName, value, key_name. The secret value is accepted on input but never returned in responses for security reasons. Required: id, keyName. - **delete_a_fireworks_ai_account_secret_by_id** — Delete a secret from a Fireworks account. Returns an empty 204 response on success. Required: id. - **get_single_fireworks_ai_account_secret_by_id** — Get a secret from a Fireworks account by id. Returns: name, keyName, value. The secret value is not returned for security reasons. Required: account_id, id. - **list_all_fireworks_ai_account_serverless_rate_limits** — List effective global rate limits for shared serverless pool deployments (fireworks-owned) as a time series. Returns rate limit time-series data (response shape not documented in source). Required: account_id. - **list_all_fireworks_ai_account_serverless_token_usages** — List serverless pool token throughput (tokens per minute) by time bucket and base model for a Fireworks account, from production usage metrics. Returns: peakTokensPerMinuteByBaseModel and averageTokensPerMinuteByBaseModel time-series sections. Required: account_id. At least one of includePeakTokensPerMinuteByBaseModel or includeAverageTokensPerMinuteByBaseModel must be true. - **list_all_fireworks_ai_account_supervised_fine_tuning_jobs** — List supervised fine-tuning jobs in a Fireworks account. Returns a collection of job objects including their name and id, with additional content-type-specific fields. Required: account_id. - **create_a_fireworks_ai_account_supervised_fine_tuning_job** — Create a supervised fine-tuning job in Fireworks. Returns the created job object including its name and id, with additional content-type-specific fields. Required: account_id. - **delete_a_fireworks_ai_account_supervised_fine_tuning_job_by_id** — Delete a supervised fine-tuning job by id. Returns an empty 204 response on success. Required: id. - **fireworks_ai_account_supervised_fine_tuning_jobs_estimate_cost** — Estimate the cost of a supervised fine-tuning job in Fireworks using the dataset, base model, and number of epochs. Returns an estimate response object. Required: account_id. - **get_single_fireworks_ai_account_supervised_fine_tuning_job_by_id** — Get a supervised fine-tuning job by id in Fireworks. Returns the job object including its dataset, baseModel, epochs, displayName, state, status, and training configuration. Required: account_id, id. - **fireworks_ai_account_supervised_fine_tuning_jobs_cancel** — Cancel a running supervised fine-tuning job in Fireworks. Returns an empty response on success. Required: account_id, supervised_fine_tuning_job_id. - **fireworks_ai_account_supervised_fine_tuning_jobs_resume** — Resume a paused supervised fine-tuning job in Fireworks. Returns the updated job object including its dataset, baseModel, epochs, displayName, state, and status. Required: account_id, supervised_fine_tuning_job_id. - **list_all_fireworks_ai_account_trails** — List trails for a Fireworks account, ordered by create_time descending by default. Returns: name, displayName, createTime, updateTime, description, createdBy, langfuseConfig, defaultModel, providerKey. Required: account_id. - **create_a_fireworks_ai_account_trail** — Create a new trail in a Fireworks account. Returns the created trail object including name, displayName, description, createTime, createdBy, defaultModel, and providerKey. Required: account_id. - **update_a_fireworks_ai_account_trail_by_id** — Update an existing trail in a Fireworks account. Returns the updated trail object including name, displayName, description, updateTime, defaultModel, and providerKey. Required: id, account_id. - **delete_a_fireworks_ai_account_trail_by_id** — Delete a trail from a Fireworks account. Returns an empty 204 response on success. Required: id, account_id. - **get_single_fireworks_ai_account_trail_by_id** — Get a single Fireworks trail by id. Returns the trail object including its displayName, description, createTime, updateTime, createdBy, defaultModel, langfuseConfig, and providerKey. Required: id. Optional: readMask to restrict returned fields. - **fireworks_ai_account_trails_get_metrics** — Get trace counts and token-usage metrics for a Fireworks trail from the tracing service. Returns a gatewayTrailMetrics object with content-type-specific metric fields. Required: trail_id. - **list_all_fireworks_ai_account_training_session_jobs** — List training session jobs for a Fireworks account. Returns an array of training session job records, each containing its name and content-type-specific attributes. Required: account_id. - **create_a_fireworks_ai_account_training_session_job** — Create a training session job that binds a Fireworks account to a shared trainer. Returns the created job including its name and content-type-specific attributes. Required: account_id. - **delete_a_fireworks_ai_account_training_session_job_by_id** — Delete a training session job and cascade-clean its child sessions and routes. Returns an empty 204 response on success. Required: account_id, id. - **get_single_fireworks_ai_account_training_session_job_by_id** — Get a single training session job by id. Returns: name, displayName, createTime, updateTime, state, status, createdBy, baseModel. Required: account_id, id. - **list_all_fireworks_ai_training_sessions** — List training sessions under a training session job in Fireworks. Returns a paginated collection of training session objects, each with id and upstream-defined attributes. Required: training_session_job_id. - **create_a_fireworks_ai_training_session** — Create a training session under a training session job in Fireworks. Returns the created training session object including its id and upstream-defined attributes. Required: account_id, training_session_job_id. - **get_single_fireworks_ai_training_session_by_id** — Get a training session by id in Fireworks. Returns the training session object including its id and upstream-defined attributes. Required: id, account_id, training_session_job_id. - **fireworks_ai_training_sessions_load_state** — Load or switch the LoRA adapter state for a training session in Fireworks. Returns the updated training session object including its id and upstream-defined attributes. Required: training_session_job_id, training_session_id. - **list_all_fireworks_ai_account_training_shapes** — List training shapes for a Fireworks account. Returns: name, displayName, description, createTime, updateTime, baseModel, deploymentShapeVersion, trainerImageTag, trainerMode, nodeCount, trainerShardingScheme, modelType, parameterCount, acceleratorType, acceleratorCount, baseModelWeightPrecision, maxSupportedContextLength. Supports AIP-160 filtering and custom ordering via filter and orderBy. - **create_a_fireworks_ai_account_training_shape** — Create a training shape in a Fireworks account. Returns the created training shape including displayName, baseModel, trainerImageTag, createTime, updateTime, and acceleratorType. Optional trainingShapeId query param assigns a custom ID (AIP-122 segment format, must start with a letter); otherwise a random ID is generated. - **update_a_fireworks_ai_account_training_shape_by_id** — Update a training shape in Fireworks. Returns the updated training shape including name, displayName, description, createTime, updateTime, baseModel, deploymentShapeVersion, trainerImageTag, trainerMode, nodeCount, trainerShardingScheme, modelType, parameterCount, acceleratorType, acceleratorCount, baseModelWeightPrecision, maxSupportedContextLength. Required: id, baseModel, trainerImageTag. - **delete_a_fireworks_ai_account_training_shape_by_id** — Delete a training shape by id. Returns an empty 204 response on success. Required: id. - **get_single_fireworks_ai_account_training_shape_by_id** — Get a Fireworks training shape by id. Returns the training shape object including name, displayName, baseModel, trainerImageTag, acceleratorType, nodeCount, createTime, and updateTime. Required: id. - **list_all_fireworks_ai_account_users** — List users in a Fireworks account. Returns: name, displayName, serviceAccount, createTime, role, email, state, status, updateTime, permissionPreset. Supports AIP-160 filter grammar and custom ordering. Required: account_id. - **create_a_fireworks_ai_account_user** — Create a new user in a Fireworks account. Returns the created user object including name, displayName, email, role, state, status, createTime, updateTime, serviceAccount, and permissionPreset. Required: account_id. - **update_a_fireworks_ai_account_user_by_id** — Update an existing user in a Fireworks account. Returns the updated user object including name, displayName, email, role, state, status, createTime, updateTime, serviceAccount, and permissionPreset. Required: id, account_id, and role in the request body. - **get_single_fireworks_ai_account_user_by_id** — Get a Fireworks account user by id. Returns the user object including name, email, role, displayName, createTime, updateTime, state, status, serviceAccount, and permissionPreset. Required: id. - **fireworks_ai_account_users_get_api_key** — Get an API key for a Fireworks account user. Returns the API key object including its name. Required: user_id. - **list_all_fireworks_ai_user_api_keys** — List API keys for a user in a Fireworks account. Returns: keyId, displayName, key, createTime, secure, email, prefix, expireTime, annotations, lastUsed. Required: account_id, user_id. - **create_a_fireworks_ai_user_api_key** — Create a new API key for a specific user in a Fireworks account. Returns the created API key object. Required: account_id, user_id. - **delete_a_fireworks_ai_user_api_key_by_id** — Delete an API key for a user in Fireworks. Returns an empty 204 response on success. Required: account_id, user_id. - **list_all_fireworks_ai_responses** — List all model responses for the authenticated Fireworks account. Returns response objects, each including its id. Max 100 per page. - **create_a_fireworks_ai_response** — Create a model response in Fireworks, optionally interacting with custom tools via the Model Context Protocol (MCP). Supports conversational continuation via previous_response_id and streaming. Returns the response object including its id. - **delete_a_fireworks_ai_response_by_id** — Delete a Fireworks model response by id. The response data is removed immediately and permanently; conversations referencing it can no longer access it. Returns a DeleteResponse confirmation object. Required: id. - **get_single_fireworks_ai_response_by_id** — List all model responses for the authenticated Fireworks account. Returns a ResponseList envelope containing the paginated results. Max 100 per page. - **create_a_fireworks_ai_completion** — Create a text completion in Fireworks for the provided prompt and parameters. Returns: id, object, created, model, choices, usage, perf_metrics, attributes. Required: prompt. - **create_a_fireworks_ai_chat_completion** — Create a chat completion in Fireworks by sending a structured conversation prompt. Returns a chat completion response object following the OpenAI Chat Completions API specification; consult the upstream Fireworks documentation for the complete request and response field structures. - **create_a_fireworks_ai_message** — Send input messages with text and/or image content to an Anthropic-compatible endpoint in Fireworks, generating the next message in a single-turn or stateless multi-turn conversation. Returns a Message object following the Anthropic Messages API format. - **get_single_fireworks_ai_generic_deployment_type_version_by_id** — Get a generic deployment type version in Fireworks. Returns: name, createTime, snapshot, updateTime, imageTag. Required: generic_deployment_type_id, id. - **delete_a_fireworks_ai_generic_deployment_type_version_by_id** — Delete a generic deployment type version in Fireworks. Returns an empty 204 response on success. Required: generic_deployment_type_id, id. - **list_all_fireworks_ai_training_shape_versions** — List training shape versions for a Fireworks training shape, ordered by create_time by default. Returns: createTime, snapshot, validated, public, latestValidated, updateTime. Required: training_shape_id. - **get_single_fireworks_ai_training_shape_version_by_id** — Get a single training shape version by account id. Returns: name, createTime, snapshot, validated, public, latestValidated, updateTime. Required: training_shape_id, id. - **update_a_fireworks_ai_training_shape_version_by_id** — Update mutable fields of a training shape version (validated, public). Returns: name, createTime, snapshot, validated, public, latestValidated, updateTime. Required: training_shape_id, id. ## How it works 1. **Link your customer's Fireworks AI account.** Use Truto's frontend SDK; we handle every OAuth and API key flow so you don't need to create the OAuth app. 2. **Authentication is automatic.** Truto refreshes tokens, stores credentials securely, and injects them into every API request. 3. **Call Truto's API to reach Fireworks AI.** The Proxy API is a 1-to-1 mapping of the Fireworks AI API. 4. **Get a unified response format.** Every response uses a single shape, with cursor-based pagination and data in the `result` field. ## Use cases - **Offer BYO Fireworks AI to your enterprise customers** — Let end users connect their own Fireworks AI accounts so inference costs, fine-tuned models, and deployments stay under their control while your SaaS orchestrates the workflows on top. - **Build a continuous fine-tuning data flywheel** — Push your users' in-product data (support tickets, code diffs, resolved conversations) into their Fireworks AI datasets and trigger recurring SFT, DPO, or reinforcement fine-tuning jobs to keep their models improving automatically. - **Power multi-tenant AI features with LoRA hot-swapping** — Deploy per-customer LoRA adapters onto shared base-model deployments so each tenant gets a personalized model without dedicating GPU clusters — ideal for AI agent builders and vertical copilots. - **Ship an MLOps cost and observability dashboard** — Sync billing usages, serverless token usage, trails, and audit logs from your users' Fireworks AI accounts into a unified panel so AI teams can track spend, latency, and ROI across models. - **Automate model promotion with evaluation gates** — Run evaluation jobs against custom evaluators after every fine-tune, and only promote checkpoints to production deployments when assertions pass — removing manual QA from the release loop. ## What you can build - **One-click Fireworks AI account connection** — Let end users link their Fireworks AI account via API key and immediately list their models, deployments, and datasets inside your product. - **In-app dataset upload and validation** — Programmatically create datasets, request upload endpoints, validate uploads, and split train/eval splits directly from your UI without users touching the Fireworks console. - **Managed fine-tuning job runner** — Estimate SFT cost, kick off supervised, DPO, or reinforcement fine-tuning jobs, stream metrics, and let users cancel or resume runs — all from your product. - **LoRA adapter marketplace and hot-swap engine** — Create, deploy, and update deployed models (LoRA adapters) on shared deployments so your customers can toggle tenant-specific behavior per request. - **Router-backed inference with fallbacks** — Provision Fireworks routers on behalf of users to load-balance across deployments and fall back to base models when a fine-tuned adapter is degraded. - **Usage and chargeback dashboards** — Pull billing summaries, billing usages, serverless token usage, and trails metrics to surface per-project spend, rate-limit headroom, and token throughput to your users. ## FAQs ### How do end users authenticate their Fireworks AI account? Users connect by providing a Fireworks AI API key, which Truto stores and injects on every request. You never handle raw credentials in your app code — Truto's connection layer manages storage, rotation, and refresh. ### Can I run inference (chat completions, messages, responses) through this integration? Yes. The integration exposes Fireworks AI's OpenAI-compatible chat completions, completions, Anthropic-compatible messages, and the responses API, so you can route generative traffic to your users' accounts and models directly. ### Does the integration cover the full fine-tuning lifecycle? It does. You can create and manage supervised fine-tuning jobs, DPO jobs, reinforcement fine-tuning jobs, and RLOR trainer jobs — including cancel, resume, cost estimation, and metrics file retrieval — plus manage datasets, checkpoints, and training sessions. ### How do I manage deployments and LoRA adapters? You get full CRUD over deployments, deployed models (LoRA adapters), deployment shapes, generic deployments, clusters, and routers. You can also scale deployments, undelete them, fetch metrics, and promote checkpoints to production. ### Can I sync billing and usage data for cost dashboards? Yes. The integration lists billing summaries, account billing usages, serverless token usages, serverless rate limits, quotas, and trails (with metrics) so you can build chargeback, forecasting, and observability views on top. ### Is evaluation and model QA supported? Fully. You can create evaluations and evaluation jobs, define and version custom evaluators (including Eval Protocol V2), preview evaluator outputs, validate assertions, and roll back or alias evaluator versions before promoting a model. ### How fresh is the data returned by list endpoints? Truto proxies requests to Fireworks AI in real time, so list and get calls reflect the current state of the user's account. For high-volume sync workloads (billing usages, audit logs, trails), you can schedule periodic pulls via Truto's sync engine.