AI QA Check tool
The AI QA Check tool sends each source/target pair to a configured LLM provider and asks it to identify quality issues — by default terminology, fluency, and accuracy. The model returns structured findings, each with a type, severity, description, and suggestion, which are recorded on the block's properties as QA issues along with the provider name and the checks performed. Blocks without a translation for the target locale are skipped.
Both a target locale and provider credentials are required, since the tool makes an external API call per block. It is the LLM-based counterpart to the rule-based QA check: it can catch meaning and fluency problems that pattern rules cannot, at the cost of latency and API usage.
Parameters
| Parameter | Type | Default | Description |
|---|---|---|---|
apiKey | string | API key for the AI provider | |
checks | string[] | Quality checks to perform (e.g. terminology fluency accuracy consistency) | |
model | string | AI model name | |
provider | string | anthropic | AI provider |
Configure these parameters interactively and copy the flow-step YAML on the Tool Reference.
Examples
Check accuracy and fluency
Restrict the model to the dimensions you care about.
provider: anthropic checks: - accuracy - fluency
Use a local Ollama model
Run the check against a locally hosted model.
provider: ollama model: llama3
Processing notes
Operates on blocks that have a translation for the target locale.
Blocks are processed in parallel by default for throughput.
Findings are stored on the block as a JSON array of QA issues.
Limitations
Requires AI provider credentials and a target locale; makes an external API call per block.
Findings are model output and reflect the chosen model's judgement; review before acting on them.
Blocks without a translation for the target locale are skipped.
← Back to the Tool Reference