Add embedding guessing game feature

[SupraSummus]

Summary

This PR introduces a new embedding guessing game feature that allows players to submit phrases and see how close their embeddings are to secret target embeddings, measured by Hamming distance.

Key Changes

Models (models.py)

• Added GuessingTarget model to represent secret embeddings that players try to match
  • Stores a phrase, its embedding, and scheduling state for embedding computation
  • Includes schedule_embedding() method to queue embedding generation via django-goals
• Added Guess model to track player submissions
  • Links a player's query phrase to a target via foreign keys
  • Enforces uniqueness constraint on (target, query) pairs
• Added _ensure_target_embedding() function to compute embeddings for guessing targets
  • Mirrors the existing _ensure_embedding() pattern for ExplorerQuery
  • Handles API failures with retry logic

Views (views.py)

• Created new guessing_router sub-router under the main router
• Implemented four view endpoints:
  • guessing_list: Shows all active guessing targets
  • guessing_detail: Displays a specific target with form and guess history
  • guessing_post: Handles form submission, creates/reuses queries and guesses
  • guess_status: HTMX endpoint for polling pending guess computations
• Added GuessForm for phrase input with validation
• Implemented _render_guess_row() helper to render individual guesses with either computed distance or HTMX polling state
• Implemented _build_guessing_detail_page() to construct the game interface with sorted computed guesses and pending items

Admin (admin.py)

• Registered GuessingTarget and Guess models in Django admin
• Added custom admin for GuessingTarget with:
  • Display of name, active status, embedding readiness, and creation date
  • Auto-scheduling of embedding computation on save
  • Read-only fields for goal and timestamp

Tests (tests.py)

• Added comprehensive test suite covering:
  • Model behavior (string representation, embedding scheduling, uniqueness)
  • Embedding computation with mocked API responses and error handling
  • View functionality (list, detail, form submission)
  • Distance calculation and display logic
  • HTMX polling behavior for pending computations
  • Edge cases (duplicate phrases, phrase length validation)

Database

• Created migration 0002_guessingtarget_guess.py with models and constraints

Notable Implementation Details

• Reuses existing ExplorerQuery model for phrase embeddings to avoid duplication
• Uses Hamming distance (bit-level differences) to measure embedding similarity
• Implements HTMX polling with 2-second intervals for real-time distance updates as embeddings compute
• Separates computed guesses (sorted by distance) from pending ones in the UI
• Follows existing patterns from the embedding explorer feature (scheduling, error handling, async computation)

https://claude.ai/code/session_01X71384QMsoYNs2yNUefF3k