BREAKING: Analysis page switched from old /insights/run to new /prompts/execute
Changes:
- Backend: Added save=true parameter to /prompts/execute
- When enabled, saves final output to ai_insights table
- Extracts content from pipeline output (last stage)
- Frontend api.js: Added save parameter to executeUnifiedPrompt()
- Frontend Analysis.jsx: Switched from api.runInsight() to api.executeUnifiedPrompt()
- Transforms new result format to match InsightCard expectations
- Pipeline outputs properly extracted and displayed
Fixes: PIPELINE_MASTER responses (old template being sent to AI)
The old /insights/run endpoint used raw template field, which for the
legacy "pipeline" prompt was literally "PIPELINE_MASTER". The new
executor properly handles stages and data processing.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- caliper_summary: use body_fat_pct (not bf_jpl)
- circ_summary: use c_chest, c_waist, c_hip (not brust, taille, huefte)
- get_latest_bf: use body_fat_pct for consistency
Fixes SQL errors when running base prompts that feed pipeline prompts.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- New placeholder: {{activity_detail}} returns formatted activity log
- Shows last 20 activities with date, type, duration, kcal, HR
- Makes activity analysis prompts work properly
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Placeholder resolver returns keys with {{ }} wrappers,
but resolve_placeholders expects clean keys.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Backend: integrate get_placeholder_example_values in execute_prompt_with_data
- Backend: now provides BOTH raw data AND processed placeholders
- Backend: unwrap Markdown-wrapped JSON (```json ... ```)
- Fixes old-style prompts that expect name, weight_trend, caliper_summary
Resolves unresolved placeholders issue.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Frontend: debug viewer now shows even when test fails
- Frontend: export button to download complete prompt config as JSON
- Backend: attach debug info to JSON validation errors
- Backend: include raw output and length in error details
Users can now debug failed prompts and export configs for analysis.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Backend: debug mode in prompt_executor with placeholder tracking
- Backend: show resolved/unresolved placeholders, final prompts, AI responses
- Frontend: test button in UnifiedPromptModal for saved prompts
- Frontend: debug output viewer with JSON preview
- Frontend: wider placeholder example fields in PlaceholderPicker
Resolves pipeline execution debugging issues.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Issue: template has NOT NULL constraint but pipeline-type prompts
don't use template (they use stages JSONB instead).
Solution: ALTER COLUMN template DROP NOT NULL before inserting
pipeline configs into ai_prompts.
Fixed Step 3 pipeline_configs migration:
- Simplified JSONB aggregation logic
- Properly scope pc alias in subqueries
- Use UNNEST with FROM clause for array expansion
Previous version had correlation issues with nested subqueries.
Migration 018:
- Add display_name column to ai_prompts
- Migrate existing prompts from hardcoded SLUG_LABELS
- Fallback: name if display_name is NULL
Backend:
- PromptCreate/Update models with display_name field
- create/update/duplicate endpoints handle display_name
- Fallback: use name if display_name not provided
Frontend:
- PromptEditModal: display_name input field
- Placeholder picker: button + dropdown with all placeholders
- Shows example values, inserts {{placeholder}} on click
- Analysis.jsx: use display_name instead of SLUG_LABELS
User-facing changes:
- Prompts now show custom display names (e.g. '🍽️ Ernährung')
- Admin can edit display names instead of hardcoded labels
- Template editor has 'Platzhalter einfügen' button
- No more hardcoded SLUG_LABELS in frontend
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Backend complete:
- Migration 017: Add category column to ai_prompts
- placeholder_resolver.py: 20+ placeholders with resolver functions
- Extended routers/prompts.py with CRUD endpoints:
* POST /api/prompts (create)
* PUT /api/prompts/:id (update)
* DELETE /api/prompts/:id (delete)
* POST /api/prompts/:id/duplicate
* PUT /api/prompts/reorder
* POST /api/prompts/preview
* GET /api/prompts/placeholders
* POST /api/prompts/generate (KI-assisted generation)
* POST /api/prompts/:id/optimize (KI analysis)
- Extended models.py with PromptCreate, PromptUpdate, PromptGenerateRequest
Frontend:
- AdminPromptsPage.jsx: Full CRUD UI with category filter, reordering
Meta-Features:
- KI generates prompts from goal description + example data
- KI analyzes and optimizes existing prompts
Next: PromptEditModal, PromptGenerator, api.js integration
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
The frontend was sending quality_filter_level to the backend, but the
Pydantic ProfileUpdate model didn't include this field, so it was
silently ignored. Profile updates never actually saved the filter.
This is why the charts didn't react to filter changes - the backend
database was never updated.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Implemented global quality_filter_level in user profiles for consistent
data filtering across all views (Dashboard, History, Charts, KI-Pipeline).
Backend changes:
- Migration 016: Add quality_filter_level column to profiles table
- quality_filter.py: Centralized helper functions for SQL filtering
- insights.py: Apply global filter in _get_profile_data()
- activity.py: Apply global filter in list_activity()
Frontend changes:
- SettingsPage.jsx: Add Datenqualität section with 4-level selector
- History.jsx: Use global quality filter from profile context
Filter levels: all, quality (good+excellent+acceptable), very_good
(good+excellent), excellent (only excellent)
Closes#31
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Notiert an 3 Stellen:
1. insights.py: TODO-Kommentar im Code
2. ROADMAP.md: Deliverable bei M0.2 (lokal, nicht im Git)
3. Gitea Issue #28: Kommentar mit Spezifikation
Zukünftig:
- GET /api/insights/run/{slug}?quality_level=quality
- 4 Stufen: all, quality, very_good, excellent
- Frontend: Dropdown wie in History.jsx
- Pipeline-Configs können Standard-Level haben
User-Request: Quality-Level-Auswahl für KI-Analysen
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Problem: Import failed with "invalid literal for int() with base 10: '37.95'"
because Apple Health exports HRV and other vitals with decimal values.
Root cause: Code used int() directly on string values with decimals.
Fix:
- Added safe_int(): parses decimals as float first, then rounds to int
- Added safe_float(): robust float parsing with error handling
- Applied to all vital value parsing: RHR, HRV, VO2 Max, SpO2, resp rate
Example: '37.95' → float(37.95) → int(38) ✓
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Problem: Errors during import were logged but not visible to user.
Changes:
- Backend: Collect error messages and return in response (first 10 errors)
- Frontend: Display error details in import result box
- UI: Red background when errors > 0, shows detailed error messages
Now users can see exactly which rows failed and why.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Problem: Import expected English column names, but German Apple Health/Omron
exports use German names with units.
Fixed:
- Apple Health: Support both English and German column names
- "Start" OR "Datum/Uhrzeit"
- "Resting Heart Rate" OR "Ruhepuls (count/min)"
- "Heart Rate Variability" OR "Herzfrequenzvariabilität (ms)"
- "VO2 Max" OR "VO2 max (ml/(kg·min))"
- "Oxygen Saturation" OR "Blutsauerstoffsättigung (%)"
- "Respiratory Rate" OR "Atemfrequenz (count/min)"
- Omron: Support column names with/without units
- "Systolisch (mmHg)" OR "Systolisch"
- "Diastolisch (mmHg)" OR "Diastolisch"
- "Puls (bpm)" OR "Puls"
- "Unregelmäßiger Herzschlag festgestellt" OR "Unregelmäßiger Herzschlag"
- "Mögliches AFib" OR "Vorhofflimmern"
Added debug logging for both imports to show detected columns.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Logs:
- CSV column names from first row
- Rows skipped due to missing date
- Rows skipped due to no vitals data
- Shows which fields were found/missing
Helps diagnose CSV format mismatches.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Problem: Import reported all entries as "updated" even when skipped
due to WHERE clause (source != 'manual')
Root cause: RETURNING returns NULL when WHERE clause prevents update,
but code counted NULL as "updated" instead of "skipped"
Fix:
- Check if result is None → skipped (WHERE prevented update)
- Check if xmax = 0 → inserted (new row)
- Otherwise → updated (existing row modified)
Affects:
- vitals_baseline.py: Apple Health import
- blood_pressure.py: Omron import
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
ModuleNotFoundError: No module named 'dateutil' beim Server-Start.
Ursache: vitals.py importiert dateutil.parser für Omron-Datumsformatierung,
aber python-dateutil fehlte in requirements.txt.
Fix: python-dateutil==2.9.0 zu requirements.txt hinzugefügt.
Nach dem Update: Docker Container neu bauen auf dem Pi:
cd /home/lars/docker/bodytrack-dev
docker compose -f docker-compose.dev-env.yml build --no-cache backend
docker compose -f docker-compose.dev-env.yml up -d
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
**Backend (insights.py):**
- Extended _get_profile_data() to fetch sleep, rest_days, vitals
- Added template variables for Sleep Module:
{{sleep_summary}}, {{sleep_detail}}, {{sleep_avg_duration}}, {{sleep_avg_quality}}
- Added template variables for Rest Days:
{{rest_days_summary}}, {{rest_days_count}}, {{rest_days_types}}
- Added template variables for Vitals:
{{vitals_summary}}, {{vitals_detail}}, {{vitals_avg_hr}}, {{vitals_avg_hrv}},
{{vitals_avg_bp}}, {{vitals_vo2_max}}
**Frontend (Analysis.jsx):**
- Added 12 new template variables to VARS list in PromptEditor
- Enables AI prompt creation for Sleep, Rest Days, and Vitals analysis
All modules now have AI evaluation support for future prompt creation.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Import endpoints for Omron blood pressure CSV (German date format)
- Import endpoints for Apple Health vitals CSV
- Import UI tab in VitalsPage with drag & drop for both sources
- German month mapping for Omron date parsing ("13 März 2026")
- Upsert logic preserves manual entries (source != 'manual')
- Import result feedback (inserted/updated/skipped/errors)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Avg blood pressure (systolic/diastolic) 7d and 30d
- Latest VO2 Max value
- Avg SpO2 7d and 30d
- Backend now provides all metrics expected by frontend
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Migration 014:
- blood_pressure_systolic/diastolic (mmHg)
- pulse (bpm) - during BP measurement
- vo2_max (ml/kg/min) - from Apple Watch
- spo2 (%) - blood oxygen saturation
- respiratory_rate (breaths/min)
- irregular_heartbeat, possible_afib (boolean flags from Omron)
- Added 'omron' to source enum
Backend:
- Updated Pydantic models (VitalsEntry, VitalsUpdate)
- Updated all SELECT queries to include new fields
- Updated INSERT/UPDATE with COALESCE for partial updates
- Validation: at least one vital must be provided
Preparation for Omron + Apple Health imports
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Backend:
- New router: vitals.py with CRUD endpoints
- GET /api/vitals (list)
- GET /api/vitals/by-date/{date}
- POST /api/vitals (upsert)
- PUT /api/vitals/{id}
- DELETE /api/vitals/{id}
- GET /api/vitals/stats (7d/30d averages, trends)
- Registered in main.py
Frontend:
- VitalsPage.jsx with manual entry form
- List with inline editing
- Stats overview (averages, trend indicators)
- Added to CaptureHub (❤️ icon)
- Route /vitals in App.jsx
API:
- Added vitals methods to api.js
v9d Phase 2d - Vitals tracking complete
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- PostgreSQL returns numeric values as Decimal objects
- psycopg2.Json() cannot serialize Decimal to JSON
- Added convert_decimals() helper function
- Converts activity_data, context, and evaluation_result before saving
Fixes: Batch evaluation errors (31 errors 'Decimal is not JSON serializable')
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
- Shows first 10 errors with activity_id, training_type_id, and error message
- Helps debug evaluation failures
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Admin endpoints for profile configuration:
- Extended TrainingTypeCreate/Update models with profile field
- Added profile column to all SELECT queries
- Profile templates for Running, Meditation, Strength Training
- Template endpoints: list, get, apply
- Profile stats endpoint (configured/unconfigured count)
New file: profile_templates.py
- TEMPLATE_RUNNING: Endurance-focused with HR zones
- TEMPLATE_MEDITATION: Mental-focused (low HR ≤ instead of ≥)
- TEMPLATE_STRENGTH: Strength-focused
API Endpoints:
- GET /api/admin/training-types/profiles/templates
- GET /api/admin/training-types/profiles/templates/{key}
- POST /api/admin/training-types/{id}/profile/apply-template
- GET /api/admin/training-types/profiles/stats
Next: Frontend Admin-UI (ProfileEditor component)
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
SQL Error: VALUES lists must all be the same length (line 130)
Cause: kcal_per_km row was missing validation_rules JSONB value
Fixed: Added validation_rules '{"min": 0, "max": 1000}'::jsonb
All 16 parameter rows now have correct 10 columns:
key, name_de, name_en, category, data_type, unit, source_field,
validation_rules, description_de, description_en
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Problem: Backend crashed on startup due to evaluation import failure
Solution: Wrap evaluation_helper import in try/except
Changes:
- Import evaluation_helper with error handling
- Add EVALUATION_AVAILABLE flag
- All evaluation calls now check flag before executing
- System remains functional even if evaluation system unavailable
This prevents backend crashes if:
- Migrations haven't run yet
- Dependencies are missing
- Import errors occur
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Automatic evaluation on activity INSERT/UPDATE:
- create_activity(): Evaluate after manual creation
- update_activity(): Re-evaluate after manual update
- import_activity_csv(): Evaluate after CSV import (INSERT + UPDATE)
- bulk_categorize_activities(): Evaluate after bulk training type assignment
All evaluation calls wrapped in try/except to prevent activity operations
from failing if evaluation encounters an error. Only activities with
training_type_id assigned are evaluated.
Phase 1.2 complete ✅
## Next Steps (Phase 2):
Admin-UI for training type profile configuration
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Problem: Creating new training types via Admin UI resulted in
'Internal Server Error' because abilities dict was passed directly
to PostgreSQL JSONB column without Json() wrapper.
Solution:
- Import Json from psycopg2.extras
- Wrap abilities_json with Json() in INSERT
- Wrap data.abilities with Json() in UPDATE
Same issue as rest_days JSONB fix (commit 7d627cf).
Closes#13
Problem: User can create multiple rest days of same type per date
(e.g., 2x Mental Rest on 2026-03-23) - makes no sense.
Solution: UNIQUE constraint on (profile_id, date, focus)
## Migration 012:
- Add focus column (extracted from rest_config JSONB)
- Populate from existing data
- Add NOT NULL constraint
- Add CHECK constraint (valid focus values)
- Add UNIQUE constraint (profile_id, date, focus)
- Add index for performance
## Backend:
- Insert focus column alongside rest_config
- Handle UniqueViolation gracefully
- User-friendly error: "Du hast bereits einen Ruhetag 'Muskelregeneration' für 23.03."
## Benefits:
- DB-level enforcement (clean)
- Fast queries (no JSONB scan)
- Clear error messages
- Prevents: 2x muscle_recovery same day
- Allows: muscle_recovery + mental_rest same day ✓
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Migration 011 removed UNIQUE constraint (profile_id, date) to allow
multiple rest days per date, but INSERT still used ON CONFLICT.
Error: psycopg2.errors.InvalidColumnReference: there is no unique or
exclusion constraint matching the ON CONFLICT specification
Solution: Remove ON CONFLICT clause, use plain INSERT.
Multiple entries per date now allowed.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Error: psycopg2.ProgrammingError: can't adapt type 'dict'
Solution: Import psycopg2.extras.Json and wrap config_dict
Changes:
- Import Json from psycopg2.extras
- Wrap config_dict with Json() in INSERT
- Wrap config_dict with Json() in UPDATE
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Problem: Photos were always getting NULL date instead of form date,
causing frontend to fallback to created timestamp (today).
Root cause: FastAPI requires Form() wrapper for form fields when
mixing with File() parameters. Without it, the date parameter was
treated as query parameter and always received empty string.
Solution:
- Import Form from fastapi
- Change date parameter from str="" to str=Form("")
- Return photo_date instead of date in response (consistency)
Now photos correctly use the date from the upload form and can be
backdated when uploading later.
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
Problem:
- Photo upload with empty date parameter (date='')
- PostgreSQL rejects empty string for DATE field
- Error: "invalid input syntax for type date: ''"
- Occurred when saving circumference entry with only photo
Fix:
- Convert empty string to NULL before INSERT
- Check: date if date and date.strip() else None
- NULL is valid for optional date field
Test case:
- Circumference entry with only photo → should work now
- Photo without date → stored with date=NULL ✓
Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
