AWS Comprehend and Braintrust Integration

Demonstration of how apps are integrated in WEXTL workflow builder[4]

Build ai workflows between AWS Comprehend and Braintrust on a visual canvas. Set up data fields sync between apps once and let data flow when events come automatically. Integration will work without code and your team will see the workflow inside WEXTL® builder.

  • 3-day trial
  • Cancel anytime
  • Up to 10 minutes quick set up
  • GDPR
  • Data Encryption
  • 2FA
  • Local Data Region

Learn about AWS Comprehend and Braintrust

Detailed info you need to know about AWS Comprehend and Braintrust integrations in WEXTL. Explore which modules apps support and how you connect them.

  • AWS Comprehend

    • AI
    • 50 searches
    • 36 actions

    Amazon natural language processing service for finding insights in text.[5]

  • Braintrust

    • AI
    • 5 triggers
    • 58 searches
    • 86 actions

    AI observability and evaluation platform for logging, experiments and prompts.[6]

How to connect AWS Comprehend to Braintrust

Set up your AWS Comprehend and Braintrust integration in 3 simple steps[2]. No coding required. Up to 10 minutes setup.

How does the integration between AWS Comprehend and Braintrust work in WEXTL?

WEXTL exposes actions and triggers for AWS Comprehend and Braintrust that help you connect apps and automate repetitive tasks[1]. A trigger is an event that happens, and an action is a step that performs when the event has occurred.

Braintrust triggersAWS Comprehend actions

Braintrust modules

  • Add a Feedback Entry to a Dataset Attaches scores, comments, tags and expected values to records that already exist in a dataset. Each item of the JSON array needs the id of its record and can score from 0 to 1.
  • Add a Feedback Entry to an Experiment Writes review feedback onto events of an experiment. Each Feedback item holds the event id plus any of scores, comment, expected, metadata, tags and source.
  • Add a Feedback Entry to Project Logs Records scores, comments or tags against logged events of a project. Send one JSON item per event id.
  • Cancel an Async SQL Query Stops a long-running SQL query that was started in the background, using its Query ID. The Status and ID of the query are returned.
  • Create a Chat Completion Sends Messages to a Model through the Braintrust AI proxy and returns the Choices of the reply. Use Extra Body to add request keys such as temperature or max tokens.
  • Create a Completion Runs a plain text Prompt against a Model through the Braintrust proxy and returns the generated Choices. Use Extra Body to add more request keys.
  • Create a Custom Column Adds a computed column to a dataset, experiment or project view, defined by an Expression under a Name. Use Subtype for child objects and Variant to pick a table variant.
  • Create a Dataset Starts a dataset under a Name inside a Project. Use Description, Tags and Metadata to describe it.
  • Create a Dataset Snapshot Freezes a dataset as it stood at one version, by pairing a Name with the Transaction ID of that version. Add a Description to label the snapshot.
  • Create a Function Saves a tool, scorer or prompt-backed function in a Project. Write Function Data as JSON, and add a Slug to give it a stable identifier.
  • Create a Group Makes a group of people for permission management. List Member Users and Member Groups to seed it, then grant it access through ACLs.
  • Create a Project Starts a project to hold experiments, logs, datasets and prompts. Use Project Group Name to place it in an existing group.
  • Create a Project Automation Sets up a log alert on a Project. Write Config as JSON with the event type, status, a BTQL filter, an interval and a webhook action.
  • Create a Project Group Collects several projects under one Name. Use Description to explain the grouping.
  • Create a Project Score Defines a score that experiments and logs can be graded with, from a Name and a Score Type. Add Categories and Config as JSON to shape the score.
  • Create a Project Tag Defines a tag within a Project for marking logs and experiments. Use Color to tell tags apart and Description to explain when to apply it.
  • Create a Prompt Saves a prompt in a Project under a Name and Slug, with versioning handled by Braintrust. Use Prompt Data for the messages and model settings, and Environment Slugs to attach environments.
  • Create a Role Defines a named role for your organization. Add Member Permissions for what it allows, and Member Roles to inherit rights from other roles.
  • Create a Service Token Issues an automation token for a Service Account in place of a personal key. Use Expires In Seconds to end its life, or leave it empty so it never expires.
  • Create a Span Iframe Embeds an external page beside trace spans, loaded from a URL and titled with a Name. Turn on Post Message to send the span data to that page.
  • Create a Temporary Credential Mints a short-lived proxy key for one Model that lasts the Lifetime in Seconds. Add Logging to have its requests recorded in a Braintrust project.
  • Create a View Saves a table view on a project, dataset or other object, with a View Type and Name. Filters and layout are stored through View Data and Options.
  • Create an ACL Grants access on one Braintrust object to a User or a Group, or through a Role. Set Permission for a single right, and Restrict Object Type to narrow the rule to child objects.
  • Create an AI Secret Stores a provider key such as a model API key under a Name in your organization. Add Type and Metadata to classify it, then read Secret Updated At later to see changes.
  • Create an Env Var Defines an environment variable on a project or other Braintrust object, picking the owner with Object Type and Object. Fill Value to store its content, or Secret Type to store it as a secret.
  • Create an Environment Registers a deployment environment under a Name and a unique Slug, which is a url-friendly identifier within the organization.
  • Create an Experiment Opens an experiment in a Project. Link it to a Dataset and Dataset Version, and set Base Experiment to compare against an earlier run.
  • Create an Ingest Key Issues an ingest key tied to one Project and labelled with a Name.
  • Create an MCP Server Registers an MCP server in a Project, reachable at its URL under a Name. Add a Description to document it.
  • Create Embeddings Converts Input text into vectors with the chosen Model, sent through the Braintrust proxy. Returns the Data, plus token Usage for the call.
  • Delete a Custom Column Removes a computed column. Add Object Type and Object to narrow the lookup, and the Subtype and Variant of the deleted column are returned.
  • Delete a Dataset Deletes a dataset from its project. The Description it had is returned along with its Name.
  • Delete a Dataset Snapshot Removes a saved snapshot of a dataset and returns the Dataset ID and Xact ID it was taken from.
  • Delete a Function Removes a function such as a tool or scorer, returning its ID and Project ID.
  • Delete a Group Deletes a group of users and returns the Organization ID it belonged to.
  • Delete a Project Deletes a project and returns its Description and Organization ID.
  • Delete a Project Automation Removes a log alert automation from its project. The User ID that created it is returned.
  • Delete a Project Group Deletes a named collection of projects. The User ID that created it is returned.
  • Delete a Project Score Removes a score definition from a project, returning its Name and Project ID.
  • Delete a Project Tag Deletes a tag defined in a project and returns the Color it had.
  • Delete a Prompt Removes a saved prompt, returning its ID, Xact ID and Project ID.
  • Delete a Role Deletes a role from your organization. The removed role is returned with its Created timestamp.
  • Delete a Service Token Revokes a service token by deleting it, returning its Preview Name and Service Account ID.
  • Delete a Span Iframe Removes an embedded span iframe from a project. The returned record shows the Deleted At time.
  • Delete a View Deletes a saved table view on an object and returns its View Type and Name.
  • Delete an ACL Deletes one access rule. The removed rule is returned with its Object Type, Object ID and User ID.
  • Delete an ACL by Match Removes whichever access rule matches the Object Type and Object, narrowed with User, Group, Permission or Role.
  • Delete an AI Secret Removes a stored AI provider secret from your organization and returns the record that was deleted.
  • Delete an API Key Deletes a personal API key and returns its Name and Preview Name to confirm which key went.
  • Delete an Env Var Deletes an environment variable and returns its Name, Object Type and Object ID.

AWS Comprehend modules

  • Attach a Resource Policy to a Model Attaches a JSON Policy to a custom classifier or recognizer version so another AWS account can import it. Use Policy Revision ID when replacing an existing policy, so a concurrent change is not overwritten. Output is the new Policy Revision ID.
  • Batch Detect Dominant Language Finds the dominant language in each of 25 texts at most, in one request. A text needs 20 characters minimum and 5 KB maximum. Results keep the order you sent.
  • Batch Detect Entities Scans up to 25 texts written in the same Language for entities. Output includes the Result List, plus an Error List for texts that failed.
  • Batch Detect Key Phrases Pulls the key phrases out of several texts in one call, all sharing a single Language. Add up to 25 Texts of 5 KB each, then read Result List in the original order.
  • Batch Detect Sentiment Rates the sentiment of every text in a list sharing one Language, 25 texts at most. Check the Error List for texts that could not be scored.
  • Batch Detect Syntax Tags the words of several texts with their parts of speech. Syntax analysis covers English, Spanish, French, German, Italian and Portuguese, which are the choices for Language.
  • Batch Detect Targeted Sentiment Judges the sentiment toward each named entity inside up to 25 texts. Targeted sentiment currently supports English only, so the Language is always English.
  • Check Whether a Text Contains PII Entities Checks up to 100 KB of Text for personal information and reports which kinds appear as Labels. PII labeling supports English and Spanish, selected with Language.
  • Classify a Document Sends text or a document to a trained classifier Endpoint and returns the Classes or Labels it assigns. Use either Text or Document Bytes, and Document Reader Settings to control how PDF and image content is read.
  • Create a Dataset Adds a training or testing dataset to a flywheel. Use Dataset Name and Input Data to give the labeled files, and Dataset Type to say whether it trains or tests. You get the Dataset ARN.
  • Create a Document Classifier Starts a custom classifier on labeled Training Data, read through the Data Access Role ARN. Use Mode to allow one class or several labels per document. You receive the Document Classifier ARN.
  • Create a Flywheel Sets up a flywheel that keeps a model improving from new data. It stores datasets and statistics at the Data Lake S3 URI. Use Active Model ARN for an existing model, or Model Type and Task Settings for a new one.
  • Create an Endpoint Provisions capacity for a trained model. Use Inference Units to size it, at 100 characters per second each and billed while it runs, and Model ARN or Flywheel to say what it serves.
  • Create an Entity Recognizer Begins training a recognizer for your own entity types from Training Data, read with the Data Access Role ARN. Use Version Name to label this version. Output includes the Entity Recognizer ARN.
  • Delete a Document Classifier Deletes a custom classifier. Only classifiers in the Trained or In Error state qualify, and not while an endpoint uses them.
  • Delete a Flywheel Removes a flywheel. It cannot be deleted while an iteration is running or an endpoint relies on it.
  • Delete a Resource Policy Deletes the policy attached to a custom model version, found by its Model ARN. Use Policy Revision ID to delete only when the policy is still at that revision.
  • Delete an Endpoint Removes a model endpoint. A model can only be deleted after all of its endpoints are gone.
  • Delete an Entity Recognizer Deletes a custom recognizer that is Trained or In Error, provided no endpoint is serving it.
  • Describe a Dataset Looks up a single dataset of a flywheel and shows its Dataset Properties.
  • Describe a Document Classification Job Reports how an asynchronous classification job is progressing, as Job Properties.
  • Describe a Document Classifier Reads one custom classifier version and its training state from the Classifier Properties.
  • Describe a Dominant Language Detection Job Looks up a language detection job that is running or finished, giving back its Job Properties.
  • Describe a Flywheel Reads the configuration and status of a flywheel, summarized as Flywheel Properties.
  • Describe a Flywheel Iteration Fetches the Iteration Properties of one training iteration of a flywheel, found with Flywheel and Iteration.
  • Describe a Key Phrases Detection Job Pulls the Job Properties of a key phrase extraction job, whether it is still going or done.
  • Describe a PII Entities Detection Job Returns Job Properties for a personal information scan, covering its progress.
  • Describe a Resource Policy Reads the policy of a custom model version, found by its Model ARN. Output includes the Policy text and its Policy Revision ID.
  • Describe a Sentiment Detection Job Checks on a sentiment analysis job and fetches its Job Properties.
  • Describe a Targeted Sentiment Detection Job Checks on a job that scores sentiment toward named entities, giving its Job Properties.
  • Describe a Topics Detection Job Gets the Job Properties of a topic modeling job that is running or finished.
  • Describe an Endpoint Fetches the current state of a model endpoint, returned as Endpoint Properties.
  • Describe an Entities Detection Job Retrieves the Job Properties for an entity detection run, which tell you whether it has finished.
  • Describe an Entity Recognizer Returns the Recognizer Properties of one custom recognizer version.
  • Describe an Events Detection Job Shows where an events detection job stands, with the Job Properties of the run.
  • Detect Dominant Language Identifies the language of a text of at least 20 characters, up to 100 KB. Output lists each candidate under Languages.
  • Detect Entities Finds entities in a piece of Text or in a document sent as Document Bytes. Use Language with the built-in model, or Custom Endpoint to find your own entity types.
  • Detect Key Phrases Extracts the main phrases from a single text under 100 KB, written in the chosen Language. Output is a list of Key Phrases.
  • Detect PII Entities Locates personal information in up to 100 KB of Text, with Language limited to English or Spanish. Output is a list of Entities.
  • Detect Sentiment Scores one text of up to 5 KB, such as a message body, review or comment from an earlier step. Output includes the Sentiment and its Sentiment Score.
  • Detect Syntax Tags each word of a text with its part of speech, within a 5 KB limit. Supported languages are English, Spanish, French, German, Italian and Portuguese. Syntax Tokens list the words.
  • Detect Targeted Sentiment Judges sentiment toward each entity in one English text, such as a product review naming several items, with a 5 KB ceiling. Entities carry their own sentiment.
  • Detect Toxic Content Scores up to 10 Text Segments for toxicity, each up to 1 KB and 10 KB in all. Toxicity detection currently works for English only, and results follow the order sent.
  • Import a Model Brings a classifier or recognizer version from another account into yours, once that account has allowed it with a resource policy. Use Model Name and Version Name to label the copy, and Model KMS Key to encrypt it. Output is the Model ARN.
  • List Datasets Lists the datasets of your flywheels, each as its own result with a Dataset Name, Dataset Type and Number of Documents. Use Filter to limit by status, type or creation date, and Flywheel to see one flywheel only.
  • List Document Classification Jobs Shows the classification jobs you have run, naming the classifier or flywheel behind each. Use Filter on the Job Name, Job Status or submit time, one at a time.
  • List Document Classifier Summaries Gives one summary per classifier name instead of per version, showing the Number of Versions and the Latest Version Status. Use Limit to cap the results of a run.
  • List Document Classifiers Returns every trained or training classifier version with its Version Name, Language Code, Status and Mode. Filter by Classifier Name, Status or when it was submitted to find one.
  • List Dominant Language Detection Jobs Retrieves your language detection jobs. Use Filter on Job Name or Job Status to find a particular run, and Limit to return fewer results.
  • List Endpoints Lists model endpoints with their Status, Current Inference Units and Model ARN. Use Filter to show only those attached to one model, in one state, or created in a date range.

Explore alternative integrations

There are other apps[9] supported by WEXTL that can be paired with AWS Comprehend and Braintrust integrations. Pick any from the list below if the current combination isn't the one you need.

Automate AWS Comprehend and Braintrust on WEXTL®

Claim your 3-day trial to build AWS Comprehend with Braintrust integration and access every WEXTL® feature.

Cancel anytime.

Frequently asked questions

Common questions about connecting AWS Comprehend and Braintrust.

How do I connect AWS Comprehend and Braintrust?

Connect your AWS Comprehend and Braintrust accounts once, pick a trigger in one app and an action in the other, then map the fields so data flows automatically between them.[3]

What can I automate between AWS Comprehend and Braintrust?

Example: when the "Watch Alerts" trigger in Braintrust fires, WEXTL can "Attach a Resource Policy to a Model" in AWS Comprehend. AWS Comprehend has 86 actions (including Attach a Resource Policy to a Model, Batch Detect Dominant Language, and Batch Detect Entities). Braintrust has 144 actions (including Add a Feedback Entry to a Dataset, Add a Feedback Entry to an Experiment, and Add a Feedback Entry to Project Logs). Add filters and transforms between steps when you need them.[1]

Can AWS Comprehend trigger actions in Braintrust, and Braintrust in AWS Comprehend?

Yes. The integration works in both directions. Use the triggers of each app to start a workflow and the actions of the other app to update it.[2]

Do I need to write code to integrate AWS Comprehend and Braintrust?

No. AWS Comprehend-to-Braintrust automations are built visually on the canvas. You can optionally add custom JavaScript with Functions when you need advanced logic.[4]

Does AWS Comprehend integrate with Braintrust in real time?

Yes. Trigger-based automations start the moment AWS Comprehend (or Braintrust) fires a new event. Webhook-backed apps run within seconds, and polling-only apps re-check on a schedule you control.[7]

How long does it take to set up a AWS Comprehend + Braintrust integration?

A first automation typically takes 5–10 minutes: connect the credential of each app once, drop a trigger + action onto the canvas, map the fields, and turn the workflow on.

Do I need a paid plan to connect AWS Comprehend with Braintrust?

You can build your AWS Comprehend + Braintrust automation on a 3-day trial with every feature unlocked. Cancel anytime.[8]

References

List of sources used to gather information about the AWS Comprehend and Braintrust integration.

  1. AWS Comprehend and Braintrust integration source code, module specs and credentials implementation.
  2. Our AWS Comprehend and Braintrust integration documentation, structured input and output parameters, scopes, and authentication methods.
  3. Official vendor API documentation for AWS Comprehend and Braintrust, used to verify field meanings, authentication flows, and anything our own specs don't fully capture.
  4. WEXTL® visual workflow builder, anonymized data from our technical support team.
  5. AWS Comprehend integration information
  6. Braintrust integration information
  7. WEXTL® Scheduling
  8. WEXTL® Pricing and Trial
  9. WEXTL® Apps Library

How we write our apps documentation

Explore how WEXTL® crafted AWS Comprehend and Braintrust integration page documentation. Used sources, the role of AI, and how the team validates the result.

Explore AWS Comprehend and Braintrust, or get started