Private Intelligence by DIGL v4.5.1aaccd96
Loading…
Prompt & features
Changes save automatically where toggles apply
About you
Tell the assistant about yourself — your role, preferences, anything useful for it to know. This is added as context, not as your own system prompt. Your name from your profile (topbar → your email) is included automatically.
System prompt
Loading…
Model
Choose the Ollama model used for chat on this device.
Knowledge base Retired
This global knowledge store has been retired — use Workspaces instead for document uploads, a persona, and durable facts scoped to you.
Web search (DDGS)
Type #search to force, #nosearch to skip.
Prompt spell-check
Each workspace gives its chats their own persona, facts and knowledge. ⚠ Any user of this box can open any workspace — don't store person-private data.
New workspace
Shown in chat headers and the workspace picker.
The role, tone and rules every chat in this workspace follows — like a standing instruction to the assistant. You can change it any time; leave blank for the standard behavior.
Short, durable truths injected into every chat here — rates, addresses, policies. ⚠ Visible to everyone using this workspace.
Chats in this workspace automatically answer from these documents (and the Efix database when attached). Upload, check status and delete on the Knowledge screen — one place for all of it.
The role, tone and rules every chat here follows — like a standing instruction to the assistant. Leave blank for the standard behavior.
Short, durable truths injected into every chat — saved when someone says "remember …" in a workspace chat, or added here. ⚠ Visible to everyone using this workspace.
Workspaces you manage are listed here. Other organization workspaces remain available in chat when you have access.
Create an organization first to create a shared workspace.
Use comma-separated email addresses. Organization managers retain management access.
Create organizations, add existing application users as members, and choose who can manage shared workspaces.
No organizations created yet.
Using “Everyone” access? First add each person here as an active organization member. “Everyone” means every active member of the selected organization.
Assets
| Name | Type | File | |
|---|---|---|---|
| No assets uploaded yet. | |||
Templates
| Name | Type | File | |
|---|---|---|---|
| No templates uploaded yet. | |||
Organization
Create shared workspaces, control access, and connect organization knowledge.
Create a workspace first
Connector knowledge belongs to a workspace. Create one, then return here to upload your organization’s documents.
Only public HTTP(S) pages on the same website are imported. Website text becomes workspace knowledge and stays subject to the same RAG and citation rules as uploaded documents.
Drop documents here
or
Efix Database — database search Not attached Manage
Ask natural-language questions against the published Efix knowledge source from a Chat using this workspace.
REST API keys
Create keys for external apps, scripts, or other machines on your network.
The local web UI on this PC does not need a key. Remote clients must send
Authorization: Bearer <key> or X-API-Key: <key>.
Rate limit: 6 requests per minute per key.
Copy this key now — it will not be shown again:
POST /chat
Authorization: Bearer <your-key>
Content-Type: application/json
{"message": "Hello", "stream": true}
OpenAI-compatible endpoint — works with any OpenAI SDK by setting
base_url to https://llm.dil.in/v1.
POST /v1/chat/completions
Authorization: Bearer <your-key>
Content-Type: application/json
{"model": "Smart",
"messages": [{"role": "system", "content": "You are an office document assistant."},
{"role": "user", "content": "Rewrite this professionally: Sales were bad."}],
"temperature": 0.3, "stream": true, "max_tokens": 1024}
model Instant | Fast | Smart (tier names, not raw model ids)
stream true → text/event-stream, ends with: data: [DONE]
GET /v1/models lists the available tiers.
Text only. This endpoint has no access to workspace documents,
saved memory, chat history or web search — send everything you
want considered in "messages". It also does not add a date, so
include one yourself if the answer depends on it.
Limit: 5 requests/min and 2 concurrent streams per key.
API usage
Last remote request per key only — not a full history. Local web UI requests are not logged here.
| Key | Last used | IP | Last prompt |
|---|---|---|---|
| Loading… | |||
Your usage
Tokens generated per day in this chat (your own usage only), and the single biggest reply each day. Precomputed — not a live recalculation.
| Date | Total tokens | Turns | Biggest reply |
|---|---|---|---|
| Loading… | |||
Egress audit — what left this machine
Every outbound web search is recorded here verbatim (after redaction) and kept forever — this is the audit record. You see your own searches plus unattributed API/automation traffic.
| When | Chat | Provider | Queries sent | Hits | |
|---|---|---|---|---|---|
| Loading… | |||||
Overview
Users
"Chat log" shows that user's real last 20 prompts + replies — the one place this dashboard isn't content-free. Every view is recorded (who viewed whose log, when).
| Role | Private web egress | Sessions | Last active | Registered | ||
|---|---|---|---|---|---|---|
| Loading… | ||||||
API keys
Every key across every user (name/prefix/owner only — never the secret or prompt text).
| Name | Prefix | Owner | Created | Last used |
|---|---|---|---|---|
| Loading… | ||||
Models
| Model | Backend | Loaded | Context | VRAM | Pinned |
|---|---|---|---|---|---|
| Loading… | |||||