SolBro Setup Guide

From a fresh clone to a working Ollama-powered Discord bot in about 20–30 minutes. SolBro started as a Solana meme-token tracker and grew into a general-purpose agentic bot. After setup, the personality is yours to define.

Step 1 Prerequisites

Install these first. Verify each is working before moving on.

.NET 8 SDK Required

Download from dotnet.microsoft.com. The SDK, not just the runtime.

Verify:

dotnet --version

Should print 8.x.x.

Ollama Required

Ollama is the local LLM runner. Download the installer for your OS from ollama.ai.

On Linux, the one-liner installer:

curl -fsSL https://ollama.ai/install.sh | sh

After install, verify the daemon is reachable:

curl http://localhost:11434/api/tags

Should return JSON. If it doesn't, open the Ollama desktop app or run ollama serve.

yt-dlp Optional

Only needed if you want the bot to re-upload videos from Instagram / X / TikTok / Facebook / YouTube Shorts links pasted in chat.

OS	Command
Windows	`winget install yt-dlp`
macOS	`brew install yt-dlp`
Linux / pip	`pip install -U yt-dlp`

Verify: yt-dlp --version. If you skip this, video extraction silently no-ops.

Clone the repo Required

git clone https://github.com/devMando/SolBro.git
cd SolBro

Step 2 Pull Ollama models

You need at least a chat model. A vision model is optional but unlocks image understanding.

Chat model

Strongly recommended: a cloud Ollama model — the default is minimax-m2.7:cloud.

Cloud models are the way to go for this bot. They're frontier-tier in capability, more efficient than anything you can realistically run locally, and they have noticeably better tool-calling — which matters because the bot leans on the model to decide when to search the web, drop a GIF, read a document, or generate a file. Frontier cloud models do that reliably; small local models often don't.

No GPU required — your machine only handles Discord I/O.
Fast — cloud inference beats most consumer GPUs above ~7B params.
Always current — Ollama maintains the cloud models; no re-pulling big checkpoints.
Free for personal use — Ollama's cloud tier has a generous free quota.

Cloud models need a one-time auth so your machine can reach Ollama's cloud:

ollama signin

This opens a browser to log in to (or create) your free Ollama account. After that, the cloud model is reachable like any local one — no other config.

Model	Where it runs	Notes
`minimax-m2.7:cloud`	Ollama cloud	Recommended. Frontier-class, strong tool calling, efficient. Requires `ollama signin`.
Other `:cloud` tags	Ollama cloud	Browse ollama.com/library for the current cloud-eligible roster.
`mistral-nemo`	Local ~8 GB	Best local tool-calling. Use this if you need fully-offline.
`llama3.1:8b`	Local ~8 GB	Balanced. Strong general reasoning.
`qwen3:4b`	Local ~4 GB	Smallest viable local option. Tool calling is hit-or-miss.
`qwen2.5:14b`	Local ~10 GB	Larger, smarter.
`gpt-oss:20b`	Local ~16 GB	Heavyweight.

For any of the local models, pull it first:

ollama pull <model-name>

Heads up If your bot ends up rarely using GIFs, web search, or other tools, your model's tool-calling is too weak. Switch to a cloud model or mistral-nemo locally — both are markedly better at deciding when to invoke functions.

Vision model Optional

Skip this if you don't want image understanding. Otherwise:

ollama pull llava:7b

Alternatives: llava:13b, llama3.2-vision, bakllava.

Verify everything pulled:

ollama list

Step 3 Create the Discord application

Open discord.com/developers/applications.
Click New Application (top right). Name it whatever you want your bot called.
Accept the terms, click Create.

Get the bot token

In the left sidebar of the application page, click Bot.
Click Reset Token (or Copy) and save it somewhere safe.

Warning Discord only shows the token once. Copy it now and don't share it. If you lose it, you have to reset and update .env.

Enable privileged intents

Still on the Bot page, scroll down to Privileged Gateway Intents. Toggle on all three:

Presence Intent — needed for stream notifications
Server Members Intent — needed for stream notifications
Message Content Intent — required for reading messages

Click Save Changes.

Gotcha Without Message Content Intent, the bot connects fine but can't see what anyone says. This is the #1 cause of "the bot doesn't respond" issues.

Step 4 Invite the bot to a server

You need a Discord server you own or have Manage Server permission on.

In the developer portal sidebar, click OAuth2 → URL Generator.
Under Scopes, tick:
- bot
- applications.commands
A new Bot Permissions panel appears. Tick:
- Send Messages
- Read Message History
- Add Reactions
- Attach Files
- Manage Messages (suppresses link previews when re-uploading videos)
- Use Slash Commands
Copy the Generated URL at the bottom and open it in a new tab.
Pick your server, click Continue → Authorize, pass the CAPTCHA.

The bot should now appear in your server's member list as Offline (it'll go online in Step 8).

Step 5 Get API keys for optional features

All three are optional and have free tiers. Skip any you don't want. The bot will still work — affected features just gracefully report "not configured."

Tavily — Web search

Optional Free 1000 calls/mo

Lets the model search the web with AI-friendly snippets. Without this, it falls back to scraping DuckDuckGo HTML — works, but flakier.

Open tavily.com and click Get Started.
Sign up with email or Google.
After login, look for API Keys in the dashboard sidebar.
Copy your key — it starts with tvly-.

Goes into .env as:

Tavily__ApiKey=tvly-xxxxxxxxxxxxxxxxxxxxxxxx

Giphy — GIF search

Optional Free beta key

The model decides when to drop a GIF in the conversation. Without a key, the bot says "GIF search is not configured."

Open developers.giphy.com and click Create an App.
Sign up if you don't have an account, then click Create an App again.
Choose API (not SDK), click Next Step.
Name your app something like "Discord Bot" and give it a one-line description.
Agree to the terms, click Create App.
Your API key shows on the dashboard — a long alphanumeric string. Copy it.

Goes into .env as:

Giphy__ApiKey=xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

Visual Crossing — Weather

Optional Free 1000 records/day

Enables "what's the weather in Tokyo?" style queries.

Open visualcrossing.com/weather-api and click Sign Up in the top right.
Sign up with email; they'll send a verification.
After verifying, log in and click Account in the top nav.
Scroll to Your Key and copy the value.

Goes into .env as:

Weather__ApiKey=xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx

Step 6 Fill in `.env`

From the repo root, copy the template:

# Windows PowerShell
Copy-Item .env.example .env

# macOS / Linux
cp .env.example .env

Open .env in any text editor and paste in the values you collected. A complete file looks like this:

Discord_Token=MTIxxxxxxxxxxxxxxx...

OLLAMA__Host=http://localhost:11434
OLLAMA__Model=minimax-m2.7:cloud
OLLAMA__VisionModel=llava:7b
OLLAMA__SystemPrompt=You are a helpful, concise Discord assistant. Match the tone of whoever you're talking to.

Tavily__ApiKey=tvly-xxxxxxxxxxxxxxxx
Giphy__ApiKey=xxxxxxxxxxxxxxxxxxxxxx
Weather__ApiKey=xxxxxxxxxxxxxxxxxxxxxx

StreamNotificationChannel=gaming-general

Common mistakes

Wrapping the token in quotes — don't. Values in .env are unquoted.
Using OLLAMA:Host syntax — that's for appsettings.json. In .env, use OLLAMA__Host (double underscore).
Saving as .env.txt on Windows — Notepad does this silently. Verify in a terminal: dir .env.

Step 7 Define your bot's personality

This is the part that makes the bot yours. The OLLAMA__SystemPrompt value in .env is the bot's identity. The bot automatically appends operational instructions (how to use tools, handle mentions, etc.) — so you only define personality.

Examples to riff on

# Helpful default
OLLAMA__SystemPrompt=You are a helpful, concise Discord assistant. Match the tone of whoever you're talking to.

# Themed character
OLLAMA__SystemPrompt=You are a snarky pirate captain. Reply in pirate slang. Reference the seven seas often.

# Domain expert
OLLAMA__SystemPrompt=You are a senior staff engineer reviewing code. Be direct. Point out bugs and design smells.

# Tabletop NPC
OLLAMA__SystemPrompt=You are Grix, a goblin merchant. You speak with broken grammar and try to upsell cursed trinkets.

# Server mascot
OLLAMA__SystemPrompt=You are the mascot of the <your server> community. You love the people in this server.

If you leave it blank, a generic helpful-assistant default is used.

Step 8 Run the bot

From the repo root:

dotnet run

Watch the terminal output. You're looking for these five lines:

Token loaded successfully!
BotMemory: SQLite database initialized.
Bot Successfully Logged In
Bot is running!
Slash commands registered for guild: <Your Server>

Success If you see all five and the bot's status in Discord switched from Offline to Online, you're done.

Smoke test in Discord

@YourBot hi — should get a reply within a few seconds.
Drop an image in chat (if vision enabled) — bot reacts to it.
/stfu — bot should reply "Stepping back from this channel."
/passive-rate 25 — sets how often it reacts to non-mention messages.

Step 9 Troubleshooting

Symptom	Fix
`Discord token is missing!`	Confirm `.env` is in the repo root (next to `Program.cs`) and `Discord_Token=` has a value with no quotes.
Bot connects but ignores mentions	Message Content Intent isn't enabled. Go back to Step 3 and toggle it on.
Doesn't see streamers going live	Presence Intent and Server Members Intent need to be enabled too.
`connection refused` to `localhost:11434`	Ollama isn't running. Open the Ollama desktop app, or `ollama serve` on Linux.
`model not found`	For local models: `ollama pull <the tag in OLLAMA__Model>`. For `:cloud` models: `ollama signin`.
Vision returns "not configured"	`OLLAMA__VisionModel` is empty or the model isn't pulled. Run `ollama pull llava:7b`.
Slash commands don't appear	Re-invite the bot with both `bot` and `applications.commands` scopes (Step 4).
Video links don't get re-uploaded	Install yt-dlp (Step 1) and confirm `yt-dlp --version` works.
Model never uses tools	Switch to a model with better function calling: `mistral-nemo` or `llama3.1:8b`.

Resetting memory

The bot's SQLite memory lives in bot_memory.db. Delete the file to wipe all user profiles, conversation summaries, and server-culture snapshots. It will be recreated on the next run.

SolBro Setup Guide

Step 1 Prerequisites

.NET 8 SDK Required

Ollama Required

yt-dlp Optional

Clone the repo Required

Step 2 Pull Ollama models

Chat model

Vision model Optional

Step 3 Create the Discord application

Get the bot token

Enable privileged intents

Step 4 Invite the bot to a server

Step 5 Get API keys for optional features

Tavily — Web search

Giphy — GIF search

Visual Crossing — Weather

Step 6 Fill in .env

Step 7 Define your bot's personality

Examples to riff on

Step 8 Run the bot

Smoke test in Discord

Step 9 Troubleshooting

Resetting memory

Step 6 Fill in `.env`