Skip to content

PromptLens v0.7.0

Latest
Latest

Choose a tag to compare

View all tags
@wildbyteai wildbyteai released this 31 Jul 03:58
v0.7.0
1b8c681

Features

  • Add a unified Prompt Compiler that shares evidence, language, Negative Prompt, variant, and platform rules across API analysis and ChatGPT Assist.
  • Add structured-output and target image-generation platform settings.
  • Add Native JSON Schema, per-template Token budgets, truncation detection, and partial-success warnings.
  • Add Anthropic Messages API tool_use support and Midjourney as a target-platform option.
  • Centralize all structured output contracts in one Prompt Schema module shared by the compiler, adapters, and validator.
  • Give Quick Copy a dedicated concise Schema that omits prompt variants and reduces its output Token budget.

Fixes

  • Preserve the marketing template type when duplicating the visual marketing template.
  • Stop reporting a fully successful analysis when required prompt variants are incomplete.
  • Stop appending API Schema instructions directly to ChatGPT Assist Markdown prompts.
  • Increment custom template versions and modification timestamps when saving edits.
  • Warn before saving a custom instruction that attempts to override the fixed format or Schema, requiring a second confirmation click.
  • Publish only the complete Simplified Chinese and English extension locales; Traditional Chinese documentation remains separate from the extension UI.
  • Prefer an explicit template type over the legacy marketing-template ID fallback.
  • Localize Chinese ChatGPT Assist section names such as English Prompt, Tags, and Negative Prompt while retaining the English professional terms in parentheses.
  • Stop a stale ChatGPT transfer retry from inserting the same instruction into the new composer after the user submits the first message.

Observability

  • Store Prompt Hash, Schema, Compiler, template version, target platform, model adapter, and validation metadata on new history records while keeping old records compatible.

Validation

  • Add a repeatable 10-image synthetic marketing benchmark for score dispersion and image-specific Negative Prompt rules without using business data.
  • Add a CI gate that requires at least 80% line and function coverage across the core runtime modules.
Assets 3
Loading