Skip to content

docs: add quality and throughput tuning guide - #261

Open
binaryaaron wants to merge 2 commits into
mainfrom
binaryaaron/docs/tuning-quality-throughput
Open

docs: add quality and throughput tuning guide#261
binaryaaron wants to merge 2 commits into
mainfrom
binaryaaron/docs/tuning-quality-throughput

Conversation

@binaryaaron

Copy link
Copy Markdown
Collaborator

Summary

  • add a quality-first guide for tuning Anonymizer
  • define quality gates, configuration freeze points, and representative benchmark design
  • document controlled throughput experiments and deployment-specific options
  • link the guide from the documentation navigation and troubleshooting workflow

Issue

No linked issue is needed for this maintainer-owned documentation improvement.

Validation

  • make docs-build
  • make format-check
  • make copyright-check
  • independent council review against the current implementation and public documentation

Public API and data safety

No public API or default behavior changes. Examples use generic placeholder data and contain no credentials or real PII.

Signed-off-by: Aaron Gonzales <aagonzales@nvidia.com>
@binaryaaron
binaryaaron requested review from a team as code owners September 8, 2026 19:26
@greptile-apps

greptile-apps Bot commented Sep 8, 2026

Copy link
Copy Markdown
Contributor

RetriggerView in GreptileConfidence Score: 5/5

The documentation-only PR appears safe to merge.

Summary

  • Defines quality gates, configuration freeze points, and representative benchmarks.
  • Describes controlled throughput experiments and deployment-specific controls.
  • Links the guide from troubleshooting and the MkDocs navigation.
  • Clarifies that Rewrite latent-entity detection is not constrained by the standard entity_labels allowlist.

Diagram

%%{init: {'theme': 'neutral'}}%%
flowchart LR
    A[Define quality gate] --> B[Establish quality configuration]
    B --> C[Freeze configuration]
    C --> D[Design representative benchmark]
    D --> E[Tune one throughput layer]
    E --> F{Quality and stability pass?}
    F -->|No| E
    F -->|Yes| G[Select smallest stable production setting]
Loading

Comment thread docs/tuning.md Outdated
Signed-off-by: Aaron Gonzales <aagonzales@nvidia.com>
Comment thread docs/tuning.md
| Detector admission | Increase the detector alias's `max_parallel_requests` gradually | GLiNER batch policy and all LLM settings |
| Validator admission | Increase validator alias limits gradually | Pool topology, chunk size, excerpt size, and model |
| Validator topology | Compare one alias with multiple equivalent aliases | Same total client-side alias limits and provider quotas |
| Validator chunking | Vary `validation_max_entities_per_call` | Validator model, excerpt size, and total client-side admission |

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Shall we suggest a range to vary it within? Could also add another column to this table with suggested values to check between or suggested values for what a "gradual" increase means for that variable

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

2 participants