Rosa Del Mar

Daily Brief

Issue 91 2026-04-01

Cost-Accounting Modularization (Feature Migration)

Issue 91 Edition 2026-04-01 4 min read
General
Sources: 1 • Confidence: High • Updated: 2026-04-12 09:59

Key takeaways

  • In datasette-llm-usage 0.2a0, features related to allowances and estimated pricing were removed.
  • datasette-llm-usage now depends on datasette-llm for model configuration.
  • If the datasette-llm-usage.log_prompts plugin setting is enabled, full prompts, responses, and tool calls can be logged to the internal llm_usage_prompt_log table.
  • The /-/llm-usage-simple-prompt page was redesigned and requires the llm-usage-simple-prompt permission to access.
  • Allowances and estimated pricing functionality is handled by datasette-llm-accountant rather than datasette-llm-usage.

Sections

Cost-Accounting Modularization (Feature Migration)

  • In datasette-llm-usage 0.2a0, features related to allowances and estimated pricing were removed.
  • Allowances and estimated pricing functionality is handled by datasette-llm-accountant rather than datasette-llm-usage.

Centralized Model Configuration Via Dependency Change

  • datasette-llm-usage now depends on datasette-llm for model configuration.

Observability Expansion With Optional Prompt/Tool Logging

  • If the datasette-llm-usage.log_prompts plugin setting is enabled, full prompts, responses, and tool calls can be logged to the internal llm_usage_prompt_log table.

Access Control Tightening On Interactive Prompt Ui

  • The /-/llm-usage-simple-prompt page was redesigned and requires the llm-usage-simple-prompt permission to access.

Unknowns

  • What is the migration path (configuration steps and data continuity) for users moving allowance/estimated pricing workflows from datasette-llm-usage to datasette-llm-accountant?
  • What versions of datasette-llm are required by datasette-llm-usage after the new dependency, and are there breaking configuration changes for existing deployments?
  • Is datasette-llm-usage.log_prompts disabled by default, and what are the recommended retention, redaction, or access-control practices for llm_usage_prompt_log?
  • What is the schema and lifecycle behavior of llm_usage_prompt_log (e.g., size growth characteristics, indexing, cleanup tooling)?
  • How is the llm-usage-simple-prompt permission intended to be assigned (roles/groups), and does the redesign introduce new capabilities beyond access gating?

Investor overlay

Read-throughs

  • Modularizing cost accounting into a separate plugin may indicate a strategy to keep core usage tracking lightweight while offering optional governance features, potentially increasing adoption in teams that need auditing and budgeting without forcing complexity on all users.
  • New dependency on datasette-llm for model configuration suggests consolidation around a single configuration layer, potentially reducing fragmentation and support burden if it simplifies deployments for usage tracking workflows.
  • Optional logging of full prompts, responses, and tool calls plus new permission gating could signal a push toward enterprise style observability and access control features, which may broaden applicability in regulated or security conscious environments.

What would confirm

  • Clear migration guidance and tooling for moving allowance and estimated pricing workflows from datasette-llm-usage to datasette-llm-accountant, including data continuity details and minimal breaking changes.
  • Documentation or release notes showing stable, compatible version requirements for datasette-llm and simplified model configuration outcomes for existing deployments.
  • Published guidance for retention, redaction, access control, and lifecycle management of llm_usage_prompt_log, indicating operational readiness for sensitive data handling.

What would kill

  • Users report painful migration, broken cost workflows, or loss of continuity after allowances and estimated pricing removal, leading to churn or reduced usage tracking adoption.
  • Dependency change introduces frequent breaking configuration issues or tight version constraints that complicate upgrades and increase support load.
  • Prompt logging defaults or schema behavior cause uncontrolled data growth, privacy incidents, or unclear access control, discouraging enabling observability features.

Sources

  1. 2026-04-01 simonwillison.net