💡

Quick overview

Content Capture

Content Capture is an enrichment module that stores a copy of the full prompt and response payload inside VeriProof, allowing reviewers and auditors to inspect the raw content of AI decisions.

Enabling content capture: Enable the content-capture enrichment in your SDK configuration and choose whether to store prompts, responses, or both.

Access control: Stored content is encrypted at rest. Access to raw content in the portal (Journey View → Payload tab) is controlled by the content:read permission, assigned via RBAC. By default, only Admin and ReviewAccess roles can view content.

Retention: Content is stored for the same retention window as decisions (default 90 days; configurable on enterprise plans). After expiry, the payload is deleted but the decision record is retained per policy.

Storage volume: Enabling content capture increases storage consumption. Monitor your storage usage under Settings → Billing.

Content Capture

Content capture controls whether VeriProof stores the raw text of prompts, model responses, and tool call inputs and outputs alongside a session trace. Captured content enables:

Time Machine — replaying exact prompts and responses in the Story panel
Review Queues — showing reviewers the full conversation that produced a decision
Evaluation Datasets — importing real production sessions as test cases with ground-truth content
Prompt Playground — running what-if experiments against the original session transcript

Without content capture, VeriProof records the structural trace (steps, latency, decision metadata, governance annotations) but the text fields are absent or summarised.

Capture Setting

Setting	What is stored	Use when
`enable_content_capture: False` (default)	No text is captured	PII-sensitive applications, highest storage efficiency
`enable_content_capture: True`	Complete input and output text from step data	Development, testing, compliance-heavy applications

Enabling content capture stores verbatim user input and model output in VeriProof’s data store. Review your data processing agreement and confirm input content does not include regulated PII before enabling in a production environment.

Setting the Capture Mode

Content capture is configured per SDK client instance, not per session. This lets you apply different policies in different deployments or services.

Python


import os
from veriproof_sdk import VeriproofClient, VeriproofClientOptions
 
# Disabled (default)
client = VeriproofClient(
    VeriproofClientOptions(
        api_key=os.environ["VERIPROOF_API_KEY"],
        application_id="my-app",
    )
)
 
# Enabled
client = VeriproofClient(
    VeriproofClientOptions(
        api_key=os.environ["VERIPROOF_API_KEY"],
        application_id="my-app",
        enable_content_capture=True,
    )
)

TypeScript


import { VeriproofClient, VeriproofSdkOptions } from '@veriproof/sdk-core';
 
// Disabled (default)
const client = new VeriproofClient(
  VeriproofSdkOptions.createDevelopment({ apiKey: process.env.VERIPROOF_API_KEY! }),
);
 
// Enabled
const client = new VeriproofClient(
  VeriproofSdkOptions.createDevelopment({
    apiKey: process.env.VERIPROOF_API_KEY!,
    enableContentCapture: true,
  }),
);

.NET


using Veriproof.Sdk;
 
// Disabled (default)
builder.Services.AddOpenTelemetry()
    .WithTraces(traces => traces
        .AddSource("*")
        .AddVeriproofTracing(options =>
        {
            options.ApiKey = Environment.GetEnvironmentVariable("VERIPROOF_API_KEY")!;
            options.ApplicationId = "my-app";
        })
    );
 
// Enabled
builder.Services.AddOpenTelemetry()
    .WithTraces(traces => traces
        .AddSource("*")
        .AddVeriproofTracing(options =>
        {
            options.ApiKey = Environment.GetEnvironmentVariable("VERIPROOF_API_KEY")!;
            options.ApplicationId = "my-app";
            options.EnableContentCapture = true;
        })
    );

Attribute Redaction

For clients with content capture enabled, you can block specific OTel attribute keys from being stored. VeriProof replaces the value of any listed attribute with [REDACTED] before transmission.

Python


client = VeriproofClient(
    VeriproofClientOptions(
        api_key=os.environ["VERIPROOF_API_KEY"],
        application_id="my-app",
        enable_content_capture=True,
        redacted_attributes=("gen_ai.input_messages", "gen_ai.output_messages"),
    )
)

TypeScript


const client = new VeriproofClient(
  VeriproofSdkOptions.createDevelopment({
    apiKey: process.env.VERIPROOF_API_KEY!,
    enableContentCapture: true,
    redactedAttributes: ['gen_ai.input_messages', 'gen_ai.output_messages'],
  }),
);

.NET


builder.Services.AddOpenTelemetry()
    .WithTraces(traces => traces
        .AddSource("*")
        .AddVeriproofTracing(options =>
        {
            options.ApiKey = Environment.GetEnvironmentVariable("VERIPROOF_API_KEY")!;
            options.EnableContentCapture = true;
            options.RedactedAttributes = ["gen_ai.input_messages", "gen_ai.output_messages"];
        })
    );

Redacted attributes are excluded before transmission. The attribute key remains visible in the trace structure, but the value is replaced with [REDACTED].

Impact on Time Machine Richness

The capture mode you choose directly affects what reviewers see in the Time Machine:

Panel	`off`	`enabled`
Story panel	Step types and timestamps only	Full verbatim transcript
DVR controls	Available	Available
Decision Snapshot	Full governance metadata	Full governance metadata
Blockchain Verification	Full	Full
Governance Context	Full	Full

If your compliance program requires human review of flagged decisions, enable content capture in your application to make the full conversation transcript visible to reviewers in Time Machine.

Storage Implications

Full capture can significantly increase storage volume. As a rough guide:

Capture mode	Storage per 1 M sessions	Notes
`off` (default)	~80 MB	Structural trace only
`enabled`	400 MB – 4 GB	Depends on average content length

Contact your account team if you need content capture storage isolated to a region-specific data store to meet data residency requirements.

See how captured content surfaces during session replay.

Time Machine

Human reviewers rely on captured content to evaluate decisions.

Review Queues

Import production sessions with full content as ground-truth test cases.

Evaluation Datasets

Language-by-language availability of built-in PII scanning.

SDK Feature Matrix

Content Capture

Content Capture

Capture Setting

Setting the Capture Mode

Python

TypeScript

.NET

Attribute Redaction

Python

TypeScript

.NET

Impact on Time Machine Richness

Storage Implications

Related Documentation