GT-ASSIST-V2
One Kenyan account for multi-model AI chat, image, video and voice — billed in KES.
The problem
A Kenyan team that wants AI in its daily work ends up juggling half a dozen separate tools — one for chat, one for images, one for video, one for transcription — each with its own USD subscription, its own login, and no way to pay in shillings. Usage is impossible to budget, and most tools cannot even be trialled without a foreign card. GT-ASSIST-V2 puts the whole set behind a single account with one credit balance and local payment rails.
What the agent does
- 1The user signs in and receives a daily free credit allowance; top-ups go through a Paystack- or Stripe-style checkout they confirm themselves.
- 2They send a chat message; the app routes the request to a model suited to the job.
- 3The reply streams back token by token, with tool calls (web search, knowledge base) surfaced inline.
- 4Media requests branch off the same credit balance: images, image edits, video animation, speech-to-text and text-to-speech.
- 5Conversations, media and the credit ledger sync to the user's account so any device resumes where they left off.
- 6Rate-limited provider calls retry with exponential backoff and jitter rather than failing the user's request.
Agent architecture
The big picture of how the agent is put together — the open-source models behind each mode, the kinds of services it talks to, and how work and memory flow. The specific tool wiring and internal API structure stay internal by design.
Models
- Modes: chat, image generation and editing, video animation, speech-to-text, text-to-speech — all from one credit balance.
- Everyday chat — DeepSeek V3.2 / V4-Flash (fast lane, 1 credit)
- Long-context work — Kimi K2.6 / K3
- Hard reasoning — GLM-5.1 / GLM-5.2 and Qwen3 235B-Thinking
- Images — Qwen-Image-2512 + FLUX.1 schnell; video — LTX 2.5 (i2v/t2v)
- Audio — Whisper (speech-to-text) and Kokoro TTS (text-to-speech)
Tools
- Model & media gateways — e.g. Chutes or OpenRouter style providers, one place to reach open chat, image, video and audio models
- Payment processors — e.g. Paystack or Stripe style, so top-ups happen in KES and every one is confirmed by the customer
- Identity — a hosted sign-in provider keeps one account across devices
- Cloud data — a managed Postgres-style sync backend keeps state consistent everywhere
- Grounding — web search and a knowledge base, so answers can be checked against sources
- Resilience — standard retry-with-backoff on transient provider errors, so hiccups stay invisible to the user
Memory
A managed cloud database keeps conversations, generated media and the per-user credit ledger in sync across devices — including the conversation branch/edit history — so any device resumes exactly where the last one left off.
Routing
The user picks a model per task, and the platform groups them by cost and depth: 1-credit lanes (DeepSeek V4-Flash, Qwen3.6-27B) for everyday chat, reasoning lanes (Kimi K3, GLM-5.2, Qwen3-235B-Thinking) for hard work, with media models routed by task type.