Use Cases
Most AI features — a support assistant, document extraction, semantic product search, an automation agent — need the same handful of pieces: a vector store, embeddings and an LLM, somewhere to run code, object storage, and the glue that ties them together. The usual approach is to assemble those from separate vendors and operate them yourself. Dodil gives you all of them as one integrated platform.
Why Dodil is faster and cheaper
Building one of these features the DIY way means standing up — then paying for and operating — a stack like this:
| What you need | Assemble it yourself | On Dodil |
|---|---|---|
| Vector search | A vector DB (Pinecone, Weaviate, Qdrant) — its own account, auth, bill | K3 Vector / VBase |
| Embeddings + LLM | An inference API (OpenAI, Cohere) — another account, auth, bill | Ignite Models (OpenAI/Cohere-compatible) |
| Running your code | Lambda / Fargate / k8s — provision, scale, pay for idle | Ignite Compute (serverless, scale-to-zero) |
| Object storage | S3 + its own account | K3 Storage |
| Ingest → chunk → embed | Glue code you write and maintain | K3 pipelines (automatic on upload) |
| Identity across all of it | Wire auth into every service | One IAM token — works everywhere (it’s even your Milvus token) |
| Monitoring, scaling, upgrades | Per service, forever | One platform |
Collapsing that stack is where the wins come from:
- Faster to ship. One platform, one SDK/CLI, one auth, and pipelines that handle ingestion for you — features that take weeks to wire up land in days. The API is typed and reflection-discoverable, so people and agents build against it without guesswork.
- Cheaper to run. One vendor and one bill instead of four to six. No integration glue to build and maintain. Serverless compute with scale-to-zero means you don’t pay for idle. No cross-service ops tax.
- Less to break. Fewer moving parts and a single identity model mean a smaller surface to secure, monitor, and keep in sync.
The savings are structural — they come from removing vendors, glue, and idle capacity, not from a pricing gimmick.
What teams build
Each use case leads with the outcome and the “why Dodil,” then links to a step-by-step example for the build.
- Customer-support assistant — answer customers and agents from your own docs and tickets.
- Document intelligence — turn PDFs and contracts into structured, queryable data.
- Product & catalog search — semantic + keyword search that understands intent.
- Agentic automation — workflows that call tools and make LLM decisions, end to end.
See also
- Examples — the step-by-step builds behind these use cases
- Platform Design — how the pieces fit (RPC-first, one auth, agent-discoverable)