Most bots forget a member the moment the conversation ends. I build ones that don't: persistent memory per user, a personality you define in JSON instead of hardcoded strings, and image generation that runs on your own GPU so there is no per-image API bill.
Fixed scope, fixed price, agreed before anything is built. Payment through Ko-fi, which accepts cards and PayPal. Half up front on anything above $400. And if your job is small, say so — small jobs are welcome and priced to match.
A bot that keeps a profile per member and carries it across sessions and restarts. Slash commands for chatting, remembering and forgetting, plus rate limiting and anti-spam built in so it survives a public server.
The same bot, plus SDXL running on your own hardware. Members type a prompt and get an image back. No per-image cost, no third-party content filter deciding what your community is allowed to generate.
A fast static site for your product or service, with an assistant that answers from your own content instead of inventing things. Loads in under a second, works on phones, no page builder required afterwards.
You have the site and the traffic; you want an assistant inside it. I retrofit retrieval over your documentation, streaming responses, per-user usage limits and a fallback path for when the model is unavailable.
A one-off script, a fix inside a bot someone else built, a small page, an integration that fits none of the four above. Tell me the scope and you get a price for exactly that job — not a package padded with things you don't need. Small budgets are fine: I'd rather quote honestly than turn you away.
You tell me what the bot should do, who it is for, and what it must never do. I come back with a fixed scope and a price before any code is written.
I ship a working version to a private channel so you can test it with real messages. You give feedback there, not in a document nobody reads.
Source code, a README that explains how to run it, and 30 days of fixes included. It runs on your server or stays on mine — your call.
A bot that forgets is usually a bot built on an in-memory dictionary. Mine persist to SQLite on every write, so a restart costs nothing and a crash costs one message.
Personality lives in a JSON file, not in forty scattered strings. You can change tone, rules and refusal behaviour without touching the code or asking me.
Image generation runs on your GPU, not on a metered API. The trade-off is real: you need a machine with a card worth using. If you don't have one, I'll tell you before you pay rather than after.
Describe it in a paragraph. If it is something I can build, you get a price and a date. If it isn't, you get a straight no and a suggestion of who could.