Perplexity API: Live AI Search Integration

Your chatbot gives outdated answers because the model doesn't know the latest news? Clients complain that the assistant doesn't see current prices. Classic LLMs are powerless: they operate with knowledge up to the training date. Perplexity API solves this by embedding web search directly into the language model's workflow. Unlike RAG with a knowledge base that needs maintenance, Perplexity eliminates the need to store and index your own corpora. In practice, this gives not only fresh answers but real cost savings: on one project infrastructure costs dropped threefold, and response time decreased from 2 seconds to 800 ms. In combination with Google Search API + a separate LLM, Perplexity is 2–3 times cheaper, and latency p99 improves by 60%.

We integrate Perplexity API into client projects: from competitor monitoring to corporate assistants. Our team has experience in NLP and computer vision, and has completed over 30 LLM integrations for businesses. Integration takes from 3 days to a week, depending on requirements for source filtering and caching. To assess your case, simply contact us — we'll select the optimal model.

Why Perplexity API for AI Search?

Perplexity is an LLM with built-in real-time web search. Unlike regular LLMs, the model automatically searches for current information and returns answers with source links. Useful for tasks requiring data freshness: news, prices, technical changes, documentation for new versions.

Case study. On one project, we replaced Google Search API with Perplexity for searching technical documentation. Result: response time dropped from 2 seconds to 800 ms, and infrastructure costs were cut by threefold. Additionally, the model provided precise citations directly in the response, simplifying verification. This substitution proved more advantageous than traditional RAG: no need to deploy a vector database or worry about data freshness.

How to Integrate Perplexity API into Your Project?

Perplexity uses an OpenAI-compatible API, so integration is straightforward for those already familiar with the OpenAI SDK.

from openai import OpenAI

client = OpenAI(
    api_key="PERPLEXITY_API_KEY",
    base_url="https://api.perplexity.ai",
)

response = client.chat.completions.create(
    model="llama-3.1-sonar-large-128k-online",
    messages=[{"role": "user", "content": "What new features are in the latest Python version?"}],
)
print(response.choices[0].message.content)

if hasattr(response, 'citations'):
    for citation in response.citations:
        print(f"Source: {citation}")

Search Configuration

response = client.chat.completions.create(
    model="llama-3.1-sonar-large-128k-online",
    messages=[{"role": "user", "content": "Latest news on GPT-5"}],
    extra_body={
        "search_domain_filter": ["openai.com", "techcrunch.com"],
        "search_recency_filter": "week",
        "return_images": False,
        "return_related_questions": True,
    }
)

Offline Models (Without Search)

response = client.chat.completions.create(
    model="llama-3.1-sonar-large-128k-chat",
    messages=[{"role": "user", "content": "Explain Dijkstra's algorithm"}],
)

Perplexity Models

What Savings Does Switching to Perplexity Bring?

Compared to using Google Search API + a separate LLM, Perplexity reduces latency and costs. In the mentioned case, infrastructure costs dropped threefold, and latency p99 improved by 60%. The model decides when to query search and when internal knowledge suffices — cutting down on unnecessary external requests.

Use Case Scenarios

• News and competitor monitoring — with recency and domain filters.
• Technical documentation check — the model immediately finds API changes.
• Research assistant — with verifiable sources that can be checked.
• Corporate search — over internal knowledge bases using search_domain_filter.

What's Included in the Work

• Setting up the Perplexity API client and test run.
• Handling citations and sources (parsing, validation, display).
• Domain filtering and recency filter configuration per task.
• Query caching to reduce latency and cost.
• Documentation and team training.

Work Process

1. Analysis — determine required data, update frequency, budget.
2. Design — select model (online/offline), configure filters.
3. Integration — connect via OpenAI SDK, write citation handler.
4. Testing — check latency (p99), answer accuracy, hallucination rate.
5. Deployment — launch in production, set up monitoring.

Timelines

• Basic integration: from 0.5 days.
• Citation and source parsing: from 1 day.
• Integration into corporate search: from 1 week.

We'll assess your project in one day. Contact us — we'll select the optimal model and integration scheme for your tasks. Guaranteed results: all citations verifiable, answers with sources. Perplexity API documentation recommends using search_domain_filter for targeted search.

Perplexity API documentation

LLM Development: Fine-Tuning, RAG, Agents, and Production Deployment

Using GPT‑4 or Claude 3.5 Sonnet through a public API is not a solution — it's just a tool. When the requirement is to "make it like ChatGPT, but on our data," there is a real engineering challenge behind it: from prompt engineering to training a 70B model on your own infrastructure. End-to-end LLM solution development is a complex stack, and we have been doing it for over 5 years. During this time, we have completed over 20 projects in generative AI: from RAG systems for legal departments to custom support agents. Where exactly your task falls depends on data, latency requirements, budget, and how critical confidentiality is.

A typical situation: the client has already tried ChatGPT, but results are unstable — sometimes accurate, sometimes hallucinating. Or they need integration into a corporate portal while complying with security policies. Let's break down each layer of the stack in detail — from RAG to production deployment.

Why Do RAG Systems Break and How to Fix It?

RAG (Retrieval-Augmented Generation) looks simple: find relevant documents, put them in context, get an answer. In practice, it fails in several places.

Chunking without overlap. Classic mistake: chunk_size=512, overlap=0. If the answer lies across two chunks, retrieval won't find either with sufficient confidence. Solution: overlap 15–25% of chunk_size, or better yet, sentence-aware splitting with spaCy or NLTK instead of naive character splitting.

Poor embedder. text-embedding-ada-002 is good for general use, but on legal or medical texts, specialized models like E5-large-v2, BGE-M3, or fine-tuned sentence-transformers on domain data outperform it. Recall@5 differences can be 15–25%.

No re-ranking. Vector search optimizes for speed, not relevance. A cross-encoder re-ranker (ms-marco-MiniLM-L-6-v2, bge-reranker-large) after initial retrieval improves top-3 accuracy with acceptable latency (+50–150ms). This is often more impactful than improving the embedding model.

Hybrid search. Dense vectors alone work poorly on exact queries: names, SKUs, codes. BM25 (sparse) finds exact matches but misses semantics. Hybrid via RRF (Reciprocal Rank Fusion) is the optimal compromise. Qdrant, Weaviate, and pgvector 0.7+ support hybrid search natively.

When to Fine-Tune Instead of Prompt Engineering?

Prompt engineering solves ~70% of LLM adaptation tasks for a domain. The remaining 30% require fine-tuning. Three indicators: the model ignores a specific output format even with detailed prompting; the task requires deep knowledge of specialized vocabulary (medicine, law); you need to significantly reduce token costs by replacing a large model with a smaller specialized one.

LoRA and QLoRA are the standard for SFT. LoRA adds trainable low-rank matrices to attention layers. A typical configuration for Llama-3 8B: r=64, lora_alpha=128, target_modules=["q_proj","v_proj","k_proj","o_proj"] yields ~0.8% trainable parameters, training on one A100 40GB. QLoRA adds 4-bit quantization (NF4) and allows fine-tuning 70B models on two A100 40GB, though speed drops by half compared to bf16.

DPO instead of RLHF. Direct Preference Optimization requires only (chosen, rejected) pairs, not scalar reward signals. DPOTrainer from the trl library (Hugging Face) implements it in a few dozen lines.

Common mistake. A dataset of 500 examples, 5 epochs, validation loss 0.8 — seems fine. But on test, the model degrades on general instructions. Cause: catastrophic forgetting. Solution: add 10–20% general instruction-following examples (Alpaca, FLAN) to the training set to preserve original capabilities.

How to Choose a Base Model: 8B or 70B?

For most tasks, fine-tuning an 8B model is sufficient. 70B is needed when deep reasoning is required or the 8B baseline does not reach the required quality even after fine-tuning. Inference cost for Llama-3 8B via vLLM on A100 is efficient; the exact cost depends on volume.

What Does PagedAttention Bring to Production?

vLLM is the first choice for serving open-source models. PagedAttention is the key technical innovation: KV-cache is managed like virtual memory in an OS, without fragmentation. This yields 2–4x higher throughput compared to naive HuggingFace Transformers inference. The vLLM documentation confirms that continuous batching and PagedAttention are the standard for high-load LLM services.

Typical numbers on A100 80GB for Llama-3 8B (bf16): 400–600 req/s, P50 latency 200–400ms, P99 latency 600–900ms at concurrency 64. For 70B on two A100 with tensor parallelism: 80–120 req/s, P99 latency 1.5–2.5s. AWQ or GPTQ quantization reduces memory consumption by 2x with quality loss within 1–3%.

Multi-Agent Systems

Agents are LLMs with access to tools: search, code execution, API calls, database interaction. Common patterns:

• ReAct (Reason + Act): the model reasons → chooses a tool → observes the result → reasons again. LangChain and LlamaIndex implement it out of the box.
• Multi-agent orchestration: multiple specialized agents with a coordinator on top. Example: coordinator → researcher (search + summarization) → coder (code generation and execution) → critic (verification). Tools: AutoGen (Microsoft), CrewAI, custom implementation on LangGraph.

In production, agent systems are non-deterministic. Essential: guardrails, step limits, logging of each step, human-in-the-loop for critical actions.

How We Work: Stages, Timeline, Deliverables

What Is Included

We deliver:

• Technical documentation (model card, configs, deployment instructions)
• Access to infrastructure (code repository, trained weights)
• 1 month of post-deployment support (consultations, bug fixes)
• Customer team training (2–3 sessions on system operation)

Timeline: basic RAG prototype — 1–2 weeks. Fine-tuning with customer data — 3–6 weeks (including data preparation). Production system with monitoring and retraining — 2–4 months. Cost is calculated individually based on data volume, model complexity, and infrastructure requirements.

We guarantee the quality of the final model with performance benchmarks and ongoing monitoring. Our engineers have hands‑on experience with dozens of production LLM systems.

Want to evaluate your project? Leave a request — we will prepare a preliminary summary within 1–2 business days. Or get a consultation on choosing the approach: RAG, fine-tuning, or hybrid — we will tell you what works best for you. Contact us to discuss your LLM development needs. Schedule a free consultation today.