Admin Dashboard
Manage users, companies, and video assignments
Create New User
All Users
| Name | Company | Role | Status | Upload | Limit | Actions |
|---|
From the archive
This box holds a subset. Ask for a video and replication brings it: the files arrive on their own, and its records follow within a minute of landing.
Bulk Assign to Company
Videos
| Video Title | Companies | Duration | Language | Actions |
|---|
Create New Company
All Companies
Pipeline Status
Current Video
Processing Errors Detected
📜 Processing Log
Manage Company Assignments
Video:
Delete Video
Edit User
Documents
Sales decks and reports, for admins only. Add a file by placing it in the server's documents/ folder.
Loading statistics...
Total Videos
--
Total Scenes
--
Total Frames
--
Total Cost
--
🌐 Language Distribution
🤖 Provider Usage
⏱️ Processing Details
Total Duration Processed
--
Average Cost per Video
--
Loading configuration...
Ollama Not Available
Local LLM options have been disabled because Ollama is not installed or running on this machine. Only cloud-based LLM providers (Claude, OpenAI) are available.
💰 Cost & ⚡ Speed Comparison
Claude (Anthropic)
OpenAI GPT-5 (NEW 🌟)
⚠️ Note: GPT-5 models use internal reasoning. Text tasks need higher token limits. See docs for details.
OpenAI GPT-4
Google Gemini
⚠️ Note: The Gemini 2.5 family is refused for API keys created after its retirement, and the Pro tier needs billing enabled. Only models this key can invoke are listed.
🏠 Neodata (Self-hosted)
Ollama (Local)
💡 Recommended Setup:
- • Frame Description (vision): Qwen3-VL 32B 🏠 (free, current setting) or Claude Sonnet for best quality
- • Scene Summaries: Mistral Small 4 🏠 (free, current setting) — prompts are per-scene, so 57K context is ample
- • Video Summary: Gemini 3.1 Flash-Lite (1M context, ~$0.06 per broadcast) or Claude Sonnet 4.6 (1M, ~$1.08). Do not use Haiku 4.5 (200K), GPT-4o-mini (128K) or any self-hosted model here — a 77-min capture is ~350K tokens, so they either truncate most of the programme or fail outright.
- • Chapter Summaries: Mistral Small 4 🏠 (free) — small per-chapter prompts, self-hosted is a good fit
- • Steps 7-8 (Entities/Titles): Qwen 3.6 35B 🏠 (free) or GPT-4o-mini
- • Best mix: Qwen3-VL 🏠 (vision) + Mistral 🏠 (steps 5/11) + Qwen 3.6 🏠 (steps 7/8) + Gemini 3.1 Flash-Lite (step 6) ≈ $0.06/video
⚡ Quick Presets
Apply a preset configuration for common use cases:
💡 Cost scales with video length, not per upload. Actual billed totals: 60 s clip ≈ $0.48 · 8 min ≈ $1.53 · 28 min film ≈ $4.02. Frame description is the largest share, so the vision model dominates this number.
Manual configuration
💡 Tip: Click a preset to automatically configure all steps. You can still customize individual settings afterwards.
🔍 Search Analytics
Monitor search queries and user feedback
Total Queries
-
Unique Users
-
Total Feedback
-
Avg Rating
-
Search Mode Distribution
Rating Distribution
Filters
Recent Search Queries
| User | Query | Mode | Results | Rating | Time | Actions |
|---|---|---|---|---|---|---|
| Loading... | ||||||
Showing 0-0 of 0
User Analytics
| User | Total Queries | Feedback Given | Avg Rating | Most Used Mode | Last Query |
|---|---|---|---|---|---|
| Loading... | |||||
Query Details
Search pipeline log
👤 Face Identity Database
Enrolled persons, embeddings, and per-video role assignments
Click the Face DB tab to load…
💬 Example Queries
Prompts shown to users on the search page. Edit text, reorder, or hide individual entries.
| Show | Order | Category | IT | EN | Actions |
|---|---|---|---|---|---|
| Loading… | |||||
Edit query
💳 Credits
User balances, provider costs, customer prices, and per-company markup overrides. Edits propagate within 60 seconds (server-side cache).
| User | Company | Remaining | Allocated | % used | Lifetime spent | Lifetime our cost | Logging | Show pill | Actions |
|---|---|---|---|---|---|---|---|---|---|
| Loading… | |||||||||
What providers charge us (per token / char / second). Update when provider prices shift.
Loading…
What each account actually does, joined to what it cost us. Cost is attributed per operation (one user action = one operation_id), not summed side by side over a date range — so each feature carries the cost it genuinely incurred.
| Account | Company | Feature | Ops | Active days | Billed | Our cost | Margin | Last used |
|---|---|---|---|---|---|---|---|---|
| Loading… | ||||||||
The inputs the rate card was calculated from. Editing one here does not re-price anything by itself — the prices in the other tabs are the derived numbers. This records what they were derived from, so they can be recalculated rather than guessed at. Enforce credit limits is the exception: it takes effect within a minute.
| Setting | Value | What it is | Last set | Action |
|---|---|---|---|---|
| Loading… | ||||
What the customer pays per minute or action. Markups (next tab) apply on top.
| Activity | Phase | Unit | Credits / unit | Description | Action |
|---|---|---|---|---|---|
| Loading… | |||||
Default markup applies to every company. Company-specific row overrides the default for that activity. Type 3x for a multiplier or 0.05 for an absolute price.
| Company | Activity | Markup | Effective | Action |
|---|---|---|---|---|
| Loading… | ||||
Set credits
Add markup
📊 User Activity
Pick a user to see their recent activity, what it cost us, what we billed them, and the margin.
| Phase | Operations | Our cost (USD) | Credits billed |
|---|---|---|---|
| Pick a user above. | |||
| When | Type | Activity | Detail | Usage | $/credits |
|---|---|---|---|---|---|
| Pick a user above. | |||||
This box
Settings that belong to this machine and replicate nowhere. Everything else under Admin edits configuration the archive owns and shares with every node.
Identity
Where this box is running
Retrieval
⚙️ System Settings
Runtime flags stored in the database. Changes take effect within 60 seconds (server-side cache).
Loading…