Update testimonial avatars to WebP and remove fake testimonials

Changes:
1. Converted all YAML files to reference .webp avatars instead of .jpg
2. Removed fabricated testimonials from all three solution pages:
   - documentation-from-video.yaml
   - sap-training-video-to-guides.yaml
   - tacit-knowledge-capture.yaml

Why: Fake testimonials are a credibility risk in B2B SaaS. Prospects
(especially SAP consultants) will verify customer names and companies.
Better to have no testimonials than obviously fake ones.

Next steps: Replace with real customer testimonials once collected.

🤖 Generated with [Claude Code](https://claude.com/claude-code)

Co-Authored-By: Claude <[email protected]>

Author: PhilippeTrounev

2020_websummit/discovery_call/index.html

@@ -1224,7 +1224,7 @@
 
     <div class="docsie-section-container" role="presentation" style="align-items: center">
             <h2 class="docsie-section-title " id="section-header-two-cfalg">To claim your <img
-                  src="https://appcdn2.docsie.io/assets/Docsie_Web_Summit.png" style="max-width: 100%%; height: 34px;" /> documentation portal design credit and a 30%% yearly discount, please book a call with our product expert.​
+                  src="https://appcdn2.docsie.io/assets/Docsie_Web_Summit.png" style="max-width: 100%; height: 34px;" /> documentation portal design credit and a 30% yearly discount, please book a call with our product expert.​
            </h2>
 
           </div>

2020_websummit/index.html

@@ -1243,7 +1243,7 @@ <h6 class="docsie-subtitle">
               </span>
               <span class="st-el">
                 <img class="st-icon" src="../assets/collision_2020/Discount.svg" />
-                <div class="cp-2"><b>30%% yearly discount</b></div>
+                <div class="cp-2"><b>30% yearly discount</b></div>
               </span>
               <span class="st-el">
                 <img class="st-icon" src="../assets/collision_2020/Money.svg" />

AGENTS.md

@@ -0,0 +1,19 @@
+# Repository Guidelines
+
+## Project Structure & Module Organization
+Core site templates live in `src/`; top-level pages sit in folders while reusable includes remain `_*.html` so the build keeps `<title>` and meta tags within the first 50 lines. Static assets belong in `assets/` and `static/` (edit SCSS in `static/scss/style.scss`). Multilingual resources live in `locale/` alongside `translation_config.yaml` and `site_config.yaml`. Blog output and caches reside in `blog/`, while Search Console tooling and reports sit in `gsc_reports/`.
+
+## Site & Blog Build Pipelines
+Install dependencies with `pip install -r requirements.txt`. Run `python main.py` to re-render English pages and `sh start.sh` to preview on `http://localhost:8081`. Use `python translate_and_build.py --quick` for rapid multilingual builds or `./build_multilingual.sh --quick` before releases to include YAML extraction and supplementary pages. Blog updates run through `python .external/BlogVi/run_blog_gen.py blog --force --all-ai`; follow with `blog/translate_all_languages.sh` (or `translate_remaining_languages.sh`) when refreshing cached translations.
+
+## Coding Style & Naming Conventions
+Adhere to PEP 8 with four-space indentation, `snake_case` helpers, pathlib, and f-strings. Keep scripts idempotent; heavy lifting belongs in modules. Jinja templates extend the appropriate `_base*.html`, use `{% trans %}` blocks, and source hreflang/canonical markup from shared partials. Static assets follow kebab-case filenames; prefix new includes with `_` so `CustomSite` treats them as partials.
+
+## Testing & Verification
+Validate changes manually. After site edits, run `python main.py` or the quick multilingual workflow and confirm generated HTML keeps SEO tags near the top, adds no 404s, and respects `site_config.yaml`. For BlogVi work, inspect `blogvi.log`, confirm `.translation_cache/<lang>/` gained expected JSON files, and spot-check regenerated `blog/articles/*/index.html`. When touching analytics, execute `python test_gsc_integration.py` and record any Search Console follow-up.
+
+## Commit & Pull Request Guidelines
+Use imperative, single-line summaries (`Fix hreflang map for glossary ru`). Bodies stay under 72 characters per line and list regenerated commands (`python main.py`, `translate_and_build.py --quick`, BlogVi scripts). PRs must call out affected locales, rebuilt directories, SEO verification steps, and any follow-up TODO items. Attach screenshots for layout or design shifts.
+
+## SEO, Localization & Analytics Checks
+Preserve canonical tags, hreflang matrices, and `<html lang>` accuracy—reuse `_hreflang.html` helpers and keep translations in sync with `site_config.yaml`. BlogVi translations already add canonical/hreflang metadata; if you filter languages, ensure the full list still renders. Monitor critical TODOs: prevent 404 regressions, keep image optimization on the roadmap, and address zero-click pages highlighted in `gsc_reports/TODO.md`. Before merging, skim newly rendered pages for structured data issues, confirm glossary/comparison modules still emit schema, and note any Search Console resubmission required.