109 Commits

Author SHA1 Message Date
Rene Fichtmueller
650de6ba9a feat: add verification evidence state model 2026-05-09 23:06:21 +02:00
Rene Fichtmueller
1af4f090f7 fix: harden TIP verification cleanup 2026-05-09 22:16:29 +02:00
Rene Fichtmueller
a43e572946 fix: advance TIP product verification robots 2026-05-09 20:19:19 +02:00
Rene Fichtmueller
fc18b00157 fix: verify copper cable semantics 2026-05-09 16:55:50 +02:00
Rene Fichtmueller
61acccf5df fix: require strict comparable transceiver evidence 2026-05-09 16:02:49 +02:00
Rene Fichtmueller
43b7250180 fix: automate equivalence research review queue 2026-05-09 07:48:11 +02:00
Rene Fichtmueller
a1a525b332 chore: sync API routes, dashboard hot-topics, MCP server, scraper package, scripts 2026-05-06 23:39:04 +02:00
Rene Fichtmueller
270bd12382 feat(dashboard): clickable LLM model selector — switch blog engine at runtime
- client.ts: BLOG_LLM_PROVIDER/OLLAMA_LLM_MODEL as mutable state (setLlmProvider/
  getLlmProvider). Reads blog-llm-settings.json on startup for persistence.
  All generate()/checkHealth()/chat() calls use dynamic provider() + llmModel()
  — no restart required for switches.

- blog.ts: POST /api/blog/llm/switch endpoint — validates provider, calls
  setLlmProvider(), writes settings file, returns previous+active state.

- index.html: all 4 model cards now clickable (cursor:pointer, hover fade).
  switchBlogLlm(provider, model) — disables cards during switch, shows
  green/red feedback toast, auto-refreshes status. SSH instruction removed.
2026-04-29 01:15:45 +02:00
Rene Fichtmueller
b5decc517f fix: hard-cap blog generation to 800-1100 words
LLM_OPTS.maxTokens 8192 → 1600, LLM_REFINE 6144 → 1800,
Step 4 master draft 8192 → 1600. Added explicit word-count
constraint to STEP4_MASTER_DRAFT prompt: HARD LIMIT 800-1100 words.

Root cause: no token ceiling → fo-blog-v6 produced 4000-5000w articles.
Generated-by label updated to fo-blog-engine-v6-length-fix.
2026-04-28 22:45:48 +02:00
Rene Fichtmueller
d7144731e0 feat(scraper): add 100+ OEM seed scrapers + tip-llm-guided inference layer
New OEM transceiver seed scrapers (94 cron-scheduled, 24/7):
- Media/Broadcast: Evertz, Grass Valley, Haivision, Viasat
- Asian Optical: FiberHome, Oplink, Accelink, Hisense Broadband
- Optical Mfrs: Lumentum, II-VI/Coherent, Source Photonics, O-Net,
  InnoLight, AOI, Sumitomo Electric, NeoPhotonics
- Industrial: GE Grid, Schweitzer, Moxa Industrial, Cisco IE,
  Phoenix Contact, Beckhoff, Omron, ABB, Siemens, Schneider, Rockwell, Belden
- Enterprise/DC: Arista, Pica8, Pluribus, DriveNets, Cisco (Meraki/Catalyst/Nexus/ASR)
- Cloud: AWS, Azure, Google Cloud, Meta
- Storage: NetApp, Pure Storage, HPE Storage, IBM Storage, Dell Storage, Hitachi Vantara
- 5G/RAN: Samsung Networks, Nokia AirScale, Ericsson RAN, Mavenir
- Security: Check Point, Barracuda, Fortinet, Palo Alto
- Telecom Optical: ADVA, PacketLight, FiberHome, Accelink, Hisense

API: tip-llm-guided inference layer (strict schema + repair-retry + safe fallback)
- POST /api/tip-llm/infer|research-plan|extract|finding|health
- Hard JSON schema enforcement, create_finding=false on empty evidence
- Confidence gate (>= 0.4), validation with consistency check

Build: added incremental=true to scraper tsconfig (OOM prevention)
Scheduler: 87 → 94 registered workers
2026-04-27 00:00:14 +02:00
Rene Fichtmueller
d5be0ba43c fix: catch non-zero exit in local train, return JSON instead of 500
- sh -lc replaced with bash to avoid dash/profile.d syntax errors
- runCommand errors now caught in local provider path
- stdout/stderr extracted from error object and returned as JSON
- No more HTTP 500 on script failure
2026-04-25 23:24:04 +02:00
Rene Fichtmueller
bfb43809a8 feat: standards audit + form factors reference
- Migration 100: adds `description` column to standards (plain-language
  DE·EN for non-technical colleagues), fills all 63 standards incl.
  complete 200G tier (SR4/DR4/FR4/LR4/ER4/CR4), copper DAC variants,
  PON family (GPON/XG-PON1/NG-PON2/25G-PON), 1.6T emerging standard

- Migration 101: new form_factors table — 20 entries covering SFP family
  (SFP→SFP112), QSFP family (QSFP+→QSFP-DD800), OSFP family (OSFP→OSFP224),
  CFP family, legacy XFP/CXP with full_name, speed, channels, status,
  supersedes chain, and bilingual plain-language descriptions

- GET /api/form-factors — new endpoint, returns all form factors with
  transceiver_count join
- GET /api/form-factors/:name — single form factor detail

Dashboard Standards tab:
  - DB description shown as subtitle in standards table rows
  - Full DE + EN description in standard detail panel
  - New Form Factors grid section with status badges, speed, channel info,
    family color coding, supersedes chain
  - openFormFactorDetail() panel with full specs + transceiver link
  - Search extended to match description + notes fields
2026-04-25 20:58:45 +02:00
Rene Fichtmueller
ff4bc34930 fix: remove IP restriction from internal-demand (Cloudflare tunnel breaks req.ip)
JWT auth + RLS is the real security boundary. IP check was blocking legitimate
dashboard users accessing via Cloudflare tunnel (X-Forwarded-For = real user IP,
not 127.0.0.1). Added /api/internal/demand/stock-analysis: demand × live stock
combined analysis with reorder_urgency, coverage_days_de, momentum_ratio.
Dashboard: new Demand × Live Stock Analyse panel with critical/low/ok/overstock chips.
2026-04-25 20:41:35 +02:00
Rene Fichtmueller
f162e03978 feat: Flexoptix internal demand intelligence + real forecast calibration
- Migration 099: flexoptix_internal_demand table with RLS + v_demand_by_speed view
- Import script: AES-256-CBC decrypt → parse 8585 SKUs → upsert with velocity class
- 279 SKUs cross-referenced to transceiver catalog; 1288 with real demand data
- New /api/internal/demand/* routes (by-speed, velocity, hype-weights, forecast-input)
  — protected by JWT auth + localhost/LAN IP restriction middleware
- Forecast engine calibrated with real Flexoptix run-rates (demand_calibrated flag)
- Dashboard: real Flexoptix Sales Velocity panel replaces DEMO DATA in Warehouse tab
  with momentum indicators, velocity class breakdown, trend arrows
- Security: data stays on private server; RLS enforces is_internal=TRUE at DB layer
2026-04-25 17:44:20 +02:00
Rene Fichtmueller
8c625ff1d2 feat(training): add TIP selflearning workflows 2026-04-25 12:21:56 +02:00
Rene Fichtmueller
9e6be570a3 feat: more switch image coverage + system health metrics + Erik monitor
switch-image-fetcher:
- Add Fortinet URL builder (11 FortiSwitch models)
- Add Quanta Cloud Technology, Allied Telesis, Ufispace, Netgear URL builders
- Fix alcatel-lucent-enterprise slug missing from URL_BUILDERS dispatcher
- Fix NVIDIA builder to skip ConnectX/BlueField adapters (not switches)
- Add aruba slug alias for hpe-aruba

health endpoint:
- Add system metrics: CPU load (1/5/15m), memory usage, disk usage
- Add load_status indicator (ok/busy/overloaded)
- Expose process RSS memory
- Used by external monitors

scripts/monitor-erik.sh:
- Cron-ready health check script for Claudi (.82) and Raspberry Pis
- Checks TIP API health endpoint (load, memory, disk, DB latency)
- Checks PM2 process state via SSH (errored/stopped detection)
- ntfy.sh push notifications (set NTFY_TOPIC env var)
- Includes systemd service + timer unit comments for auto-install
2026-04-21 00:31:43 +02:00
Rene Fichtmueller
a2492d833b feat: Flexoptix order section per switch + reject generic/logo images 2026-04-20 23:31:36 +02:00
Rene Fichtmueller
c0cd0dc1ca feat: compatibility panel — verification_method, competitor prices, spec-match collapsible
- API: getCompatibleTransceivers() returns verification_method, orders vendor_compat first
- Dashboard: Flexoptix section splits vendor-tested vs spec-match (collapsed)
- Dashboard: Competitor section shows vendor-tested with prices, spec-match as chips
2026-04-20 22:52:49 +02:00
Rene Fichtmueller
0fb4850dfa fix: price-comparison SKU lookup — wrong column refs (so.stock_level, search_url_template) 2026-04-19 00:12:18 +02:00
Rene Fichtmueller
446ac667b0 feat: side-by-side competitor comparison + fix 1.6T speed_gbps
- Fix OSFP-DR8-1.6T-FL and OSFP-2FR4-1.6T-FL: speed_gbps was 200, now 1600
  → FS.com 1.6T products now correctly match as comparables for Flexoptix O.1316T.C.05.M
- API: extend comparable price query to return comp_form_factor, comp_speed_gbps,
  comp_reach_meters, comp_reach_label, comp_fiber_type, comp_wavelengths
- Dashboard: replace plain comparable price row with side-by-side spec comparison card
  showing Flexoptix vs. competitor: Form Factor, Speed, Reach, Fiber, Wavelengths
  with color coding (green=match, orange=mismatch) and savings badge (−45% günstiger)
2026-04-18 21:51:41 +02:00
Rene Fichtmueller
62d97a783c feat: add claude-code LLM provider + update dashboard to fo-blog-v5
- client.ts: add claude-code provider routing BLOG_LLM_PROVIDER=claude-code
  to claude-bridge (flat-rate, no API billing via Claude Code subscription)
- checkHealth() now pings /health on claude-bridge for real availability check
- Default OLLAMA_LLM_MODEL changed from qwen2.5:14b to fo-blog-v5
- Dashboard: add claude-code card (EMPFOHLEN), rename fo-blog-v3 → fo-blog-v5
- loadBlogLLMStatus() handles all 3 providers: claude-code/anthropic/ollama
- Grid expanded from 3 to 4 columns to accommodate new card
- ecosystem.config.js + .env on Erik: OLLAMA_LLM_MODEL=fo-blog-v5 confirmed
2026-04-18 20:45:14 +02:00
Rene Fichtmueller
1c8dec52c9 feat: Price Comparison dashboard + Eoptolink OEM scraper
- Add public /api/price-comparison API (summary, top-50, per-SKU detail)
  — no auth required, 3 Express routes, DISTINCT ON latest-price logic
- Add '💲 Price Comparison' dashboard tab: stat cards, form-factor
  breakdown, top-50 SKU table (clickable rows → SKU detail), per-vendor
  price + stock + spread% lookup panel
- Add Eoptolink OEM catalog scraper (93 product-solution pages,
  part-number regex EOLO-*/EOLQ-* etc., no prices, seeds transceivers
  table as manufacturer entries)
- Register scrape:catalog:eoptolink in scheduler: schedule every 4h
  (40 */4 * * *), lazy-import worker, added to known-jobs array
2026-04-18 01:02:08 +02:00
Rene Fichtmueller
e9fcda2811 feat: wire finder.ts + switch-docs + Ollama LLM tools to MCP server
MCP Server (packages/mcp-server/src/index.ts):
- Register registerSwitchDocTools (switch-docs.ts) — switch documentation lookup
- Register finderTools dynamically (finder.ts) — find_flexoptix_for_switch, get_competitor_alerts
- Add analyze_market_with_llm tool: qwen2.5:14b via Ollama, enriched with live hype cycle + pricing + news
- Add generate_blog_post tool: fo-blog-v5 (fine-tuned) with qwen2.5:14b fallback, enriched with live pricing data
- OLLAMA_BASE_URL env var (default: https://ollama.fichtmueller.org)

Also includes scraper improvements (ascentoptics, atgbics, gbics, skylane, ebay-enricher),
API route updates (blog, blog-sll, health, hot-topics, transceivers, queries),
and dashboard hot-topics refresh.
2026-04-18 00:21:58 +02:00
Rene Fichtmueller
b88a6e28cf feat: /api/hype-cycle/analysis endpoint — DB-backed Bass-fitted results from hype_cycle_analysis table 2026-04-18 00:11:08 +02:00
Rene Fichtmueller
861243ea3f feat: stock confidence badges, multi-vendor price comparison, expanded Cisco TMG + Juniper HCT
Stock API & Dashboard:
- /api/stock/summary: vendor_breakdown adds avg_confidence, currencies, conf_per_warehouse/aggregated/boolean
- /api/stock/summary: new price_comparison endpoint (multi-vendor SKUs, min/max/avg price)
- /api/stock/summary: totals adds multi_vendor_skus count
- Dashboard: 6th stat card (Multi-Vendor SKUs), confidence badge column (🟢 L3 / 🟡 L2 /  L1)
- Dashboard: price comparison table with vendor-by-vendor price breakdown
- Dashboard: subtitle updated to include QSFPTEK + NADDOD
- Dashboard: top sellers link to product URLs

Cisco TMG improvements:
- Added 5 new platform families: 8000 Series, NCS5500, NCS540, NCS560, NCS1000
- Per-device query strategy: iterates all switch model IDs from family filter
  instead of getting only 1 switch per family → 58 switches per N9300 run
- Graceful error handling per device with rate limiting (1s between requests)

Juniper HCT: ran manually → 475 Juniper-brand transceivers seeded
2026-04-17 23:33:31 +02:00
Rene Fichtmueller
5b35b2b8be feat(scraper+api): warehouse stock data pipeline — FS.com v2, SmartOptics v2, Stock API
Scraper changes:
- fs-com.ts v2: Playwright stealth patches + www.fs.com/de/ URL fix (de.fs.com DNS NXDOMAIN).
  Extracts DE-Lager, Global-Lager, Nachlieferung, units_sold, compatible_brands, price_net.
  Mac-side runner (run-fs-scraper-mac.sh) via SSH tunnel for residential IP access.
  Fast-fail connectivity check on datacenter IPs that are blocked by Cloudflare.
- smartoptics.ts v2: WooCommerce REST API fallback + 8 catalog categories + relative URL fix.
  Was finding only 8 products, now discovers 18+ with multi-category crawl.

DB layer:
- db.ts: add upsertStockObservation() — writes 10 new stock_observations columns
  (warehouse_de_qty, warehouse_global_qty, backorder_qty, units_sold, compatible_brands,
  price_net, product_url, delivery dates) with dedup check.

API:
- routes/stock.ts: GET /api/stock, /api/stock/summary, /api/stock/:id
  Warehouse breakdowns per transceiver/vendor with top-sellers and vendor summary.
- routes/review.ts: equivalence review queue (approve/reject/bulk-approve).
- index.ts: register /api/stock and /api/review routes.

Dashboard:
- index.html: 🏭 Stock tab with stat cards (DE-Lager, Global-Lager, Nachlieferung totals),
  top-sellers table, vendor breakdown, recently-restocked events, part-number lookup.

SQL migrations:
- 034: blog-review-tag, 035: price-observations is_anomalous, 036: transceiver-equivalences.
2026-04-17 10:45:59 +02:00
Rene Fichtmueller
cddc92c9d2 feat: TIP audit fixes — Qdrant init, switches columns, verification fix, crawler live status, demo data badges
- Migration 032: add system_type, is_linecard, chassis_model, slot_type, flexbox_* to switches table
- Migration 032: fix compute_transceiver_verification() to count seed data as details_verified (100% now)
- Migration 032: add is_demo_data flag to reorder_signals, abc_classification, market_intelligence, stock_snapshots
- Cisco 8000: insert 8812, 8818, 8800-LC-36FH, 8800-LC-48H with correct vendor slug 'cisco'
- API: add /api/scrapers/jobs endpoint exposing pg-boss job queue (active/recent/queues)
- Dashboard: live job queue panel in Crawler Intelligence tab (active jobs + recent 4h completions)
- Dashboard: DEMO DATA badge now uses is_demo_data column (was checking wrong field is_demo)
- Blog engine: configured fo-blog-v3-qwen7b fine-tuned model via tip-api ecosystem.config.js
- Qdrant: all 6 collections created, seeded (2135 products, 29 FAQs, 39 news, 20 troubleshooting)
2026-04-09 20:29:46 +02:00
Rene Fichtmueller
cf75eee8ad feat: linecard system support, Cisco 8000 accuracy, price anomaly detection
API/finder:
- Add modular chassis support: sibling linecards fetched when is_linecard=true
- Add chassis linecards when system_type=modular
- Extend switch response: system_type, is_linecard, chassis_model, slot_type,
  flexbox_compat_mode, flexbox_notes, description, switching_capacity_tbps,
  total_ports, category, lifecycle_status, features, use_cases, linecards[]

API/transceivers:
- Filter price_observations with COALESCE(is_anomalous, false) = false
  (direct prices + comparable market prices)

Scraper/db:
- Add PRICE_BOUNDS map (per form-factor min/max USD sanity bounds)
- Add isPriceAnomalous() — marks DB price_observations as is_anomalous=true
- Add competitor_verified flag: set true when valid competitor price stored
- upsertPriceObservation: skip prices outside sanity bounds, set competitor_verified

Scraper/hash:
- contentHash() now accepts Record<string,unknown> | string (union type)
  to support both structured objects and legacy string callers

Scrapers (skylane, tscom, wiitek):
- Fix contentHash() call signature: pass objects not JSON.stringify strings
- Fix wiitek: remove invalid 'name' param, fix t.id → transceiverId

Migrations:
- Add is_anomalous, competitor_verified, competitor_verified_at,
  image_primary columns
- Recreate sync_fully_verified trigger to include competitor_verified
- Add is_linecard, chassis_model, system_type, slot_type,
  flexbox_compat_mode, flexbox_notes to switches table
2026-04-09 09:06:22 +02:00
Rene Fichtmueller
51af249361 Merge remote-tracking branch 'github/main'
# Conflicts:
#	packages/api/src/llm/fo-blog-pipeline.ts
#	packages/api/src/routes/blog.ts
#	packages/scraper/src/scheduler.ts
#	packages/scraper/src/scrapers/fs-com.ts
#	packages/scraper/src/scrapers/gbics.ts
2026-04-06 18:03:36 +02:00
Rene Fichtmueller
3928755c60 fix: correct verified badge, comparable pricing, and clickable product images
- Reset details_verified=false for 298 products where reach_label is empty (DB migration)
- Runtime check in dashboard: dVer requires non-empty reach_label regardless of DB flag
- comparable price query: treat reach_meters=0 same as NULL so 800G OSFP products
  find FS.com equivalent prices (was blocked by reach_meters=0 != NULL shortcircuit)
- Product image area now fully clickable with vendor link overlay when product_page_url exists
- Clear wrong image for O.Czz8HG.z.R (was showing unrelated OSFP product image)
2026-04-06 10:24:39 +02:00
Rene Fichtmueller
4acf293690 fix(llm): checkHealth uses key presence check, not live API call
Live Anthropic API call during health check causes 429 when the pipeline
is actively running, blocking all subsequent regenerate requests.
2026-04-06 04:07:21 +02:00
Rene Fichtmueller
72033ff5c5 fix(blog): fix claudeQueue deadlock from recursive 429 retry
The generateClaude() function was recursively calling itself inside
enqueueClaude(), creating a circular Promise dependency that permanently
deadlocked the claudeQueue. Any 429 rate-limit response would poison
the queue, blocking all future Claude API calls until server restart.

Fixes:
- Split retries into claudeApiCall() which is called from enqueueClaude
  (not re-entering the queue on retry = no circular dependency)
- Max 3 retries with increasing backoff (10s/30s/60s)
- Add resetClaudeQueue() exported function
- Add 15-minute auto-reset stall detection to enqueueClaude
- Expose resetClaudeQueue in POST /api/blog/llm/reset-queue endpoint
- Fix merge conflict markers in index.ts (duplicate scraperRouter import)
2026-04-06 02:51:28 +02:00
Rene Fichtmueller
a35817c96d fix: preserve user-provided title in blog generation + price floor validation
- blog/generate now uses caller title when provided; falls back to template
- Migration 027: hard price floor by speed class in verification function
  (no medians, no estimates — only real prices above minimum thresholds)
- Deleted 474 obviously wrong price observations (shipping costs scraped as prices)
2026-04-06 01:14:37 +02:00
Rene Fichtmueller
b6928265bf fix: serialize Claude API calls via queue to prevent 429 rate-limit spam
Tier-1 Anthropic API has 40K TPM — with ~20K tokens per pipeline step,
concurrent calls immediately hit the limit. enqueueClaude() serializes
all generateClaude() calls so only one runs at a time, eliminating
the flood of 429-retry-429-retry loops.
2026-04-06 00:57:03 +02:00
Rene Fichtmueller
cf04549b1b feat: add Anthropic Claude provider to blog LLM client
- Auto-routes to Claude API when BLOG_LLM_PROVIDER=anthropic + ANTHROPIC_API_KEY set
- Fallback to Ollama queue when key not present
- Add rate-limit retry (429 → 10s backoff) for Claude API
- Add STEP_TECHNICAL_SANITY, STEP_SELF_HEAL, STEP_TITLE_CONTRACT_CHECK prompts
- Fix STEP_LINKEDIN_POST angle-specific hooks, remove Gold Reference repetition
2026-04-06 00:21:48 +02:00
Rene Fichtmueller
c43e1f881a fix(blog): hard story blacklist in STEP4 + LinkedIn — ban 2AM/dirty connector/lab-vs-prod stories
10 specific story patterns banned directly in draft prompt.
LinkedIn banned hooks: 'Everything looks fine', 'CRC creeping', 'Same optics same setup'.
Title Contract now injected into STEP4 as binding constraint.
2026-04-05 23:55:56 +02:00
Rene Fichtmueller
80435f8e07 feat(blog): Post to Ghost + LinkedIn buttons in dashboard
- 'Post on blog.fichtmueller.org' → publishes via Ghost Admin API
- 'Post on LinkedIn' → modal with text + copy + open LinkedIn
- Ghost integration: TIP Blog Engine (JWT auth, mobiledoc format)
2026-04-05 23:33:58 +02:00
Rene Fichtmueller
3913372f10 feat(blog): Title Contract + Technical Sanity Check + Self-Heal + angle-aware LinkedIn generator
Pipeline now has 21 steps:
- STEP0: Title Contract binds LLM to headline promise
- STEP19: Technical Sanity Check (optical engineering accuracy)
- STEP20: Self-Heal (auto-fix technical errors preserving tone)
- STEP21: Title Contract Verification (final gate check)
- LinkedIn generator is now angle-aware (no more default Physical Layer hook)
2026-04-05 23:11:16 +02:00
Rene Fichtmueller
8e0eda6c41 fix(blog): anti-repetition engine — 6 angle types, forbidden structures, existing article context injection 2026-04-05 22:47:15 +02:00
Rene Fichtmueller
438225cf7c fix(blog): raise word target to 1200-1600, fix power-budget false positive in validateArticle 2026-04-05 20:49:22 +02:00
Rene Fichtmueller
7c7d9f7b51 fix: make /api/hot-topics public — dashboard fetch has no auth token 2026-04-05 12:07:44 +02:00
Rene Fichtmueller
e6d042f827 fix: resolve merge conflict in index.ts + add untracked blog-sll, news, sql migration 2026-04-05 11:51:07 +02:00
root
5d8768b43b fix: mount blogSllRouter + scraperRouter — SLL and Crawler Intelligence routes were missing 2026-04-05 09:50:30 +00:00
Rene Fichtmueller
e4dfd2a2db feat(blog): AEM/APM pipeline steps + SLL context builder + LinkedIn v2 prompts 2026-04-05 01:26:09 +02:00
Rene Fichtmueller
95a8aa8552 fix: include linkedin_post in GET /api/blog response for SLL matching 2026-04-05 01:24:52 +02:00
Rene Fichtmueller
931588fffd fix(verification): 100% Verified Badge war dramatisch zu großzügig
KERNPROBLEME BEHOBEN:
1. ATGBICS part_number = URL slug statt echte OEM-Nummer
   extractOemPartNumber() entfernt -r-compatible-transceiver-* Suffix
   + trailing Vendor-Namen (nokia, cisco, juniper, ...)
   Ergebnis: 3he16564aa-nokia-r-compatible-transceiver-... → 3HE16564AA

2. reach_label = '' (leer) wurde als details_verified akzeptiert
   IS NOT NULL erlaubt leere Strings → Fix: AND reach_label != ''

3. details_verified = true trotz garbled part_number
   Neue Kriterien: NOT ILIKE '%-compatible-transceiver%'
                   NOT ILIKE '%-r-compatible%'

4. data_confidence Werte falsch in Funktion ('scraped_unverified' etc)
   Echte Werte: low/medium/high/garbage → NOT IN ('garbage','unknown')

ERGEBNIS nach recompute_all_verification():
  fully_verified: 3.654 → 581 (Badge war 6x übertrieben)
  details_verified: inflated → 1.075 (korrekt)

ATGBICS Scraper:
  - extractOemPartNumber() für collection und product detail pages
  - detectReach() jetzt auch auf URL-slug (120km im slug → reach_label)

Price Anomaly Detection:
  - API: price_anomaly field wenn max/min ratio ≥ 10x
  - Dashboard: ⚠ Preisanomalie Banner mit Ratio + EUR Range

SQL 025: Part number cleanup (30 records), reach from slug (12 records)
2026-04-04 15:41:57 +02:00
Rene Fichtmueller
1e789f67eb fix(scrapers): Flexoptix Catalog zeigt 0 records statt 963
SCRAPERS list used 'flexoptix-catalog' as DB lookup key but vendors.slug
is 'flexoptix' — no match → 0 records shown.

Fix: added dbSlug override field to SCRAPERS entries; lookup now uses
dbSlug || name so flexoptix-catalog/vendors/supported all map to
the correct 'flexoptix' slug in sourceMap.
2026-04-04 15:26:04 +02:00
Rene Fichtmueller
fea0b0fb66 feat: blog engine v5 — Auto-Kill Layer, 16-step pipeline, longer content
Upgrades FO Blog Pipeline from 14 to 16 steps:
- NEW Step 8d: Auto-Kill Layer v1.0 (10 systematic categories A-J)
- NEW Step 15: Auto-Kill Scoring (cleanliness, narrative, non-AI, relevance)
- Updated banned phrases from Gold-standard editorial feedback
- Soft Delete List for conditional phrases
- Auto-Kill categories: spec blocks, formulas, section leakage,
  generic transitions, repeated concepts, SKU mentions, false authority,
  over-explained basics, whitepaper tone, fake precision

Content length changes per user feedback:
- Blog target: 1,200-2,000 words (was 700-1,000) — thorough and detailed
- LinkedIn target: 2,000-2,800 chars (was 350-600) — use maximum length
- Reduction pass: 25-30% cut (was 15-25%) — remove weak, keep depth
2026-04-04 11:02:45 +02:00
Rene Fichtmueller
ede4f5b966 feat: blog engine v3 — 8-stage pipeline with Auto-Kill Layer
Complete rewrite of blog prompts and pipeline based on editorial
Gold-standard feedback. Replaces 3-pass system with 8-stage pipeline:

1. Master generation (narrative voice, no spec dumps)
2. Narrative Control (kill visible structure, enforce flow)
3. Auto-Kill Layer (remove AI phrases, spec residue, repetition)
4. Reduction Engine (cut 40% — keep strongest ideas only)
5. Depth pass (add specifics where vague, no spec dumps)
6. Quality Control (hard delete list validation)
7. Procurement layer (optional, sales audience)
8. LinkedIn post generation (new)

Key changes:
- System prompt rewritten with Hard Delete List (29 banned phrases)
- Soft Delete List for conditional phrases
- Auto-Kill categories A-J (spec blocks, formulas, whitepaper tone, etc.)
- Master prompts enforce continuous narrative, no section headings
- Word count targets reduced (800-1200 instead of 1500+)
- Scoring pass added (cleanliness, narrative, non-AI feel, relevance)
- LinkedIn companion post auto-generated
- Context data injection reduced (fewer items, no dump instructions)
2026-04-04 10:52:31 +02:00
Rene Fichtmueller
c509251109 feat(blog): Spec dump hard fail + Gold Standards 6 + LinkedIn v2
- System prompt: SPEC DUMP ABSOLUTE HARD FAIL block (before FORMAT rules)
  TX/RX tables, multi-optic comparison blocks, repeated sections = hard fail
  Behavioral prose rule: "what actually happens" not "what the spec says"

- STEP9 QA: check 12a SPEC DUMP — removes datasheet blocks, flags
  duplicate sections (e.g. "fiber types" twice), spec-heavy intros

- Gold Standard 6: 400G/800G deep dive corrected (8.8→10/10)
  zero spec tables, pure behavioral narrative, 3 core ideas max,
  ending is reframe not checklist

- LinkedIn Gold Example 2: sharper short format (346 chars vs 700)
  reframe hook, short beats without bullet markers, no emoji, 4 hashtags

- STEP_LINKEDIN_POST: rewritten with new gold format
  optimal 350-600 chars, beat rhythm, no bullet markers, gold example inline

- WRONG PATTERNS: +7 new entries (spec dump, duplicate section,
  LinkedIn bullet list, LinkedIn "excited to share" hook, LinkedIn >800 chars)
2026-04-04 09:32:01 +02:00