gora.
EN
EnglishРусскийLietuvių
claude api · rag · llm integration · remote

Claude API Integration Developer

Gora is an independent full-stack developer who adds Claude to existing products and builds new ones around it: streaming answers over WebSocket, prompt caching, retrieval over your own content with cited sources, fallbacks to another provider and per-request cost tracking. Several of these systems run in production, from a multi-model Telegram assistant to a RAG chatbot product.

Discuss a Claude integrationView relevant work ↓
  • 12 models from four providers in one production assistant
  • Claude with prompt caching and streaming over WebSocket
  • $43.3 in API fees for 1,783 generations

Concrete kinds of work, each with a system I run in production.

  • Claude in an existing back end

    The Claude API with streaming and prompt caching, added to an existing NestJS or FastAPI service without rewriting it.

  • RAG over your content

    Retrieval with pgvector and voyage-3-large embeddings; Claude answers from your pages or documents and cites the sources. chatbot.hiregora.com is my own product built this way, with Claude Haiku 4.5 by default and Sonnet 4.6 on the Pro plan.

  • Streaming answers

    Replies stream to the browser over WebSocket as they are generated, so users read the first words immediately instead of waiting for the full answer.

  • Model choice per task

    Smaller and larger Claude models for different jobs: Haiku as the default in the chatbot product, Sonnet on the higher plan, and Opus, Sonnet and Haiku side by side in a multi-model Telegram assistant. I evaluate on your own examples before choosing.

  • Fallbacks and escalation

    If one provider fails, another one answers. When confidence is low, the question goes to a human instead of a guess.

  • Claude as one step in a pipeline

    In GoraGen, Claude picks the model for a video job and Gemini scores the result on four criteria before the pipeline decides whether to regenerate.

  • Usage and cost control

    Per-request token logs and plan-level limits enforced in the database. In GoraBot, 1,783 generations across 12 models used 5.4M input and 1.06M output tokens.

  • Guardrails around the model

    Rules before a reply is sent, confidence thresholds, content moderation and automated scoring steps, because nobody can promise a model is never wrong.

Three production systems with their numbers. Two link to full case studies.

chatbot.hiregora.com — Claude RAG chatbot for your site

My own product, in open beta: give it a URL and it builds a knowledge base and answers visitors.

What was done
Claude Haiku 4.5 answers with retrieval over pgvector and streams each reply over WebSocket. A sitemap-aware crawler builds the knowledge base from a URL, and the dashboard lists top questions and the topics the bot could not answer.
Problem solved
Let a website answer visitors from its own content, with cited sources, without a separate developer or agency.
Technology
  • Claude Haiku 4.5
  • voyage-3-large embeddings
  • FastAPI + Celery
  • PostgreSQL 16 + pgvector (HNSW)
  • Redis
  • Preact widget
In numbers
  • About 2 minutes from a URL to the first dialog
  • Claude Haiku 4.5 by default, Sonnet 4.6 on the Pro plan
  • 1024-dimension voyage-3-large embeddings; answers cite their sources

Open the product →

GoraBot — Claude next to three other providers

One Telegram chat with 12 models from Anthropic, OpenAI, Google and xAI.

What was done
Claude Opus, Sonnet and Haiku run next to GPT, Gemini and Grok in one chat: chains where four models work in sequence, debates with a judge model, project contexts, voice replies and per-plan limits enforced in the database.
Problem solved
Compare and combine different models without losing context, under one rate limit and one billing logic.
Technology
  • python-telegram-bot 21
  • Anthropic, OpenAI, Google, xAI SDKs
  • SQLite
  • FastAPI
In numbers
  • 12 models in one chat from four providers
  • 56 users and 1,783 generations in production on the publication date
  • 5.4M input and 1.06M output tokens, $43.3 in API costs

Read the case study →

GoraGen — Claude inside a video pipeline

Multi-tenant platform for AI video: self-serve LoRA training, ComfyUI on rented GPUs and storyboards.

What was done
Claude picks a model for each video job, a rented GPU instance runs ComfyUI and Gemini scores the result on four criteria; the system then decides whether to regenerate. A control plane with leader leases recovers after crashes.
Problem solved
Generate usable video without a human reviewing every attempt, while keeping expensive GPU workers from running when nobody needs them.
Technology
  • FastAPI
  • PostgreSQL 16
  • Anthropic Claude
  • Google Gemini 2.5 Pro
  • ComfyUI
  • Vast.ai
  • Docker Compose
In numbers
  • Gemini scores each result on 4 criteria before the pipeline decides
  • 61 SQL migrations and four apps behind one Caddy
  • Idempotency middleware on 5 mutating endpoints

Read the case study →

LLM APIs
Anthropic Claude (Haiku, Sonnet, Opus) with prompt caching · OpenAI GPT · Google Gemini · xAI Grok
Retrieval
pgvector (HNSW) · voyage-3-large embeddings · ChromaDB
Back end
FastAPI · NestJS · PostgreSQL · Redis · WebSocket streaming
Operations
Docker Compose · systemd · Sentry · GitHub Actions

A short path from the problem to something running, with the model's limits spelled out up front.

  1. 01 · First message

    Write to me on Telegram. In your own words, even if you don't know the technical terms. I care more about understanding what you want to build than what you call it.

  2. 02 · Reading between the lines

    Within a day or two I understand the task and give you an estimate of timeline and budget. If I see it isn't my case — I say so right away, no hedging.

  3. 03 · Lock-in

    Price and timeline — in writing. 50% upfront. After that things change only by mutual agreement, and rarely.

  4. 04 · Work

    I show interim results at my own pace. At the end — handover with documentation and access. First month of support is included by default.

  • Direct communication with the developer who writes and runs the system, with no agency layer.
  • I keep a clear line between what the model does and what stays ordinary deterministic code.
  • Production experience with Claude and three other providers at once, not a single-vendor demo.
  • I plug into existing back ends and products instead of proposing a rewrite.
  • Architecture, back end, front end and deployment from one person.
  • Working remotely with international clients; NDA on request.

I work remotely with companies in different countries, communicating in English or Russian. Most integrations are done over Telegram, email and video calls.

You keep control of the provider accounts: API keys, billing and data stay with you, and I sign an NDA on request.

Two separate costs apply. Development is scoped per project: it depends on the integration surface, whether retrieval over your documents is involved, how many channels the assistant must cover and how strict the reliability requirements are. I send a written estimate of timeline and budget, usually within a day or two of your first message.

Usage is billed by the model provider to your account. As one real data point, GoraBot's 1,783 generations across 12 models used 5.4M input and 1.06M output tokens and cost $43.3. Cost per request depends heavily on the model, which is why I log tokens per request from the start.

How much does a Claude API integration cost?
Development is quoted per project, because it depends on what the integration touches, whether retrieval over your content is needed and how many channels it must cover. Usage is billed separately by Anthropic to your account. Send a short description and I reply with a written estimate within a day or two.
Can you add Claude to my existing product?
Yes. I add Claude calls to an existing NestJS or FastAPI back end, with streaming and prompt caching where they help, without rewriting the product.
Do you use prompt caching and streaming?
Yes, where they pay off: prompt caching to cut cost and latency when the same context is sent repeatedly, and streaming over WebSocket so users see the answer as it is written.
Which Claude model should I use?
It depends on the task, latency and cost. In my chatbot product Haiku 4.5 is the default and Sonnet 4.6 is used on the higher plan. I test the candidates on your own examples before recommending one.
Can Claude answer from my documents or website?
Yes. I build retrieval with pgvector and voyage-3-large embeddings, and Claude answers from the retrieved passages and cites the sources. chatbot.hiregora.com is my own product built on this: give it a URL and it builds the knowledge base.
Can I use Claude together with OpenAI or Gemini?
Yes. Several of my systems route between providers and fall back automatically if one is unavailable; GoraBot puts 12 models from four providers behind a single chat.
How do you keep Claude's answers reliable?
With checks around the model: rules before a reply is sent, confidence thresholds with escalation to a human, fallbacks between providers, content moderation and, in generative pipelines, an automated scoring step. These controls limit the damage of a wrong answer; nobody can promise a model is never wrong.
Who owns the system and the API keys?
You do. Code goes into your GitHub repository, provider accounts and API keys stay with you, and I hand over documentation and access at the end. I sign an NDA on request and the first month of support is included.

Describe the task in your own words. I respond within 24 hours, usually the same day on weekdays.

Elsewhere: GitHub·Telegram channel·Reddit·X