11 KiB
Glossary
Quick lookup for General Bots terms. If you're lost, start here.
A
Auto-Bootstrap - The magical first run that installs everything automatically. PostgreSQL, cache, storage - all configured without you lifting a finger.
Argon2 - The Fort Knox of password hashing. Makes brute-force attacks computationally infeasible. We use it everywhere passwords are stored.
B
BASIC - Yes, that programming language from 1964. We brought it back because TALK "Hello" beats await ctx.send() any day. Powers all conversation scripts. Used since 2018 in production systems.
Bot Package - A folder ending in .gbai containing everything to run a bot. Scripts, documents, config. That's it. Copy folder = deploy bot.
BotSession - The conversation between user and bot. Remembers everything - who you are, what you said, where you left off. Persists to database, cached for speed.
Bootstrap - Initial setup process that automatically installs all dependencies. Runs on first launch, creates configuration, sets up database, configures services.
C
Collection - A folder of documents in .gbkb/ that becomes searchable knowledge. Drop PDFs in policies/, bot answers policy questions. Zero configuration.
Context - What the LLM knows right now. Includes conversation history, active knowledge bases, loaded tools. Limited by token window (usually 4-8k tokens).
config.csv - The only config file you need. Simple key-value pairs in CSV format. Opens in Excel. Lives in .gbot/ folder.
D
Dialog - A .bas script defining conversation flow. Not complex state machines - just simple BASIC code like TALK "Hi" and answer = HEAR.
Drive - Built-in S3-compatible storage (SeaweedFS). Stores documents, templates, uploads. Auto-installed during bootstrap. No AWS account needed.
E
Embedding - Text converted to numbers for similarity search. "dog" and "puppy" have similar embeddings. BGE model by default, replaceable with any GGUF model.
Event Handler - BASIC code that runs when something happens. ON "login" runs at login. ON "email" when email arrives. ON "0 8 * * *" at 8 AM daily.
G
.gbai - "General Bot AI" folder. Contains entire bot. Example: support.gbai/ becomes bot at /support. No manifest files, no build process.
.gbdialog - Subfolder with BASIC scripts. Must contain start.bas as entry point. Tools go in tools/ subdirectory.
.gbkb - "Knowledge Base" subfolder. Each subdirectory becomes a searchable collection. PDFs, Word docs, text files - all automatically indexed.
.gbot - Configuration subfolder. Contains single config.csv file with bot settings. Missing values use defaults.
.gbtheme - Optional UI customization. CSS files, images, HTML templates. Most bots don't need this.
.gbdrive - File storage configuration. Maps to Drive (S3-compatible) buckets for document management.
General Bots - The enterprise conversational AI platform. Combines LLMs with structured dialogs, knowledge bases, and multi-channel support.
H
HEAR - BASIC keyword to get user input. answer = HEAR "What's your name?" waits for response.
Hot Reload - Changes to BASIC scripts apply immediately. No restart needed. Edit, save, test - instant feedback loop.
I
Installer - Component that auto-configures services. Manages PostgreSQL, Cache, Drive, Directory Service, LLM servers. Everything runs locally.
Intent - What the user wants to do. Detected from natural language. "I want to reset my password" → password_reset intent.
K
Knowledge Base - Documents that become searchable answers. PDFs, Word files, web pages. Automatically chunked, embedded, and indexed for semantic search.
L
LLM - Large Language Model. The AI brain. Default uses llama.cpp with GGUF models. Supports OpenAI, Anthropic, local models.
Local-First - Everything runs on your machine. No cloud dependencies. Database, storage, LLM - all local. Privacy by default.
M
Memory - Bot and user memory storage. SET BOT MEMORY "key", "value" persists data. GET BOT MEMORY "key" retrieves it.
Multi-Channel - Same bot works everywhere. WhatsApp, Teams, Web, API. Write once, deploy anywhere.
O
OIDC - OpenID Connect authentication. Handled by Directory Service (Zitadel). No passwords stored in General Bots.
P
Package Manager - System that installs bot packages. Drop .gbai folder, it's automatically loaded. Remove folder, bot stops.
PostgreSQL - The database. Stores users, sessions, messages, memory. Auto-installed, auto-configured. Just works.
Pragmatismo - Company behind General Bots. Brazilian software consultancy. Building bots since 2016.
Q
Qdrant - Vector database for semantic search. Optional component for large-scale knowledge bases. Faster than PostgreSQL pgvector for millions of documents.
R
REPL - Read-Eval-Print Loop. Interactive BASIC console for testing. Type commands, see results immediately.
S
Semantic Search - Finding by meaning, not keywords. "How do I change my password?" finds "reset credentials" documentation.
Session - Active conversation state. Tracks user, bot, context, memory. Expires after inactivity. Stored in PostgreSQL, cached in memory.
SET CONTEXT - BASIC command to add information to LLM context. SET CONTEXT "User is premium customer" influences all responses.
T
TALK - BASIC keyword for bot output. TALK "Hello!" sends message to user. Supports markdown, images, cards.
Token - Unit of text for LLMs. Roughly 4 characters. Context windows measured in tokens. GPT-4: 8k tokens. Local models: 4k typically.
Tool - Function the bot can call. Defined in BASIC with parameters. PARAM "city" then weather = GET "weather" calls weather API.
U
USE KB - BASIC command to activate knowledge base. USE KB "policies" makes policy documents searchable in conversation.
USE TOOL - Activate a tool for LLM to use. USE TOOL "calculator" lets bot do math.
V
Valkey - Redis-compatible cache. Stores sessions, temporary data. Faster than database for frequently accessed data.
Vector - Mathematical representation of text meaning. Used for semantic search. Created by embedding models.
W
WebSocket - Real-time connection for chat. Enables streaming responses, live updates. No polling needed.
Workflow - Sequence of dialog steps. Login → Verify → Action → Confirm. Defined in BASIC, no complex orchestration.
Z
Zitadel - Current Directory Service implementation. Handles authentication, users, permissions. Can be replaced with Keycloak or other OIDC providers.
H
HEAR - BASIC keyword to get user input. name = HEAR waits for user to type, stores response in variable.
Hot Reload - Change scripts while bot runs. Edit file, bot uses new version immediately. No restart needed (unless changing config).
K
Knowledge Base (KB) - Documents the bot can search. Organized in folders under .gbkb/. Use with USE KB "foldername" in scripts.
Keyword - BASIC command like TALK, HEAR, LLM. About 40 total. All caps by convention, not requirement.
L
LLM - Large Language Model (ChatGPT, Claude, Llama). The AI brain that powers natural conversation understanding.
Local Mode - Run everything on your machine. LLM runs locally (llama.cpp), no internet required. Slower but private.
M
MCP - Model Context Protocol. Standard format for defining tools that LLMs can call. Alternative to OpenAI function format.
Memory - Two types: Session (temporary, per conversation) and Bot (permanent, across all sessions). SET for session, SET BOT MEMORY for permanent.
P
Package Manager - Built-in system that installs/manages components. Handles PostgreSQL, cache, storage, vector DB, LLM server. All automatic.
PARAM - Defines tool parameters. PARAM name, email, date means tool needs these three inputs. LLM collects them automatically.
PostgreSQL - The database. Stores users, sessions, messages, bot config. Auto-installed, auto-configured, auto-migrated.
Q
Qdrant - Vector database for semantic search. Stores document embeddings. Finds similar content even with different words. Optional but recommended.
R
Rhai - The scripting engine powering BASIC interpreter. Rust-based, sandboxed, safe. You never see it directly.
S
Script - A .bas file with BASIC code. start.bas runs first. Can call others with RUN "other.bas".
Semantic Search - Finding by meaning, not keywords. "vacation policy" finds "time off guidelines". Powered by embeddings and vector similarity.
Session - See BotSession. The container for a conversation.
Session Token - Random string identifying a session. Stored in browser localStorage for web, passed as header for API. Unguessable for security.
T
TALK - BASIC keyword to send message. TALK "Hello world" displays that text to user.
Template - Example bot in templates/ folder. Copy, modify, deploy. default.gbai is minimal starter, announcements.gbai shows advanced features.
Tool - A .bas file the LLM can call. Has PARAM definitions and DESCRIPTION. Put in tools/ folder. AI figures out when to use it.
Token - Unit of text for LLMs. Roughly 4 characters. GPT-3.5 handles 4k tokens, GPT-4 handles 8k-128k. Includes prompt + response.
U
USE KB - BASIC command to activate knowledge base. USE KB "policies" makes policies folder searchable for that session.
USE TOOL - BASIC command to enable a tool. USE TOOL "send-email" lets LLM send emails when appropriate.
V
Valkey - Redis-compatible cache (fork after Redis license change). Stores session data for fast access. Auto-installed during bootstrap.
Vector - Mathematical representation of meaning. "Cat" might be [0.2, 0.8, -0.3, ...]. Similar meanings have similar vectors.
Vector Collection - See Collection. Folder of documents converted to vectors for semantic search.
W
WebSocket - Real-time connection for chat. Enables instant messaging without polling. Path: /ws on same port as HTTP.
Symbols
.bas - File extension for BASIC scripts. Plain text files with BASIC code.
.csv - Configuration format. Simple, editable in Excel, no JSON parsing needed.
.env - Environment variables file. Auto-generated during bootstrap with credentials and settings.
Common Confusions
"Do I need containers?" - No. BotServer installs everything directly or in LXC containers.
"What database?" - PostgreSQL, auto-installed, auto-configured.
"What about scaling?" - Single server handles 1000+ concurrent users. Scale by running multiple instances.
"Is BASIC really BASIC?" - Inspired by BASIC, not strict implementation. Simpler, focused on conversations.
"Can I use TypeScript/Python/etc?" - BASIC is used for conversation logic. However, you can integrate with any language through APIs. See API documentation for REST endpoints and integration options.
"Is it production-ready?" - Yes. Used in production since 2016 (earlier versions), current Rust version since 2023.
