Valyu¶

What it is¶

Valyu is an AI-native search API that provides agents with access to both the open web and licensed, high-signal proprietary data sources.

What problem it solves¶

It allows agents to search beyond just the current web, providing structured, high-accuracy results from datasets like PubMed, SEC filings, clinical trials, patents, arXiv, and financial data through a single, natural-language-enabled API.

Where it fits in the stack¶

AI Assistants & Knowledge / Understand (Aggregators). It acts as a high-signal search engine that feeds real-time context and deep research data to LLMs and agents.

Typical use cases¶

Deep Research: Running complex queries that require cross-referencing web search with research papers (arXiv) or patents.
Financial Analysis: Extracting real-time market data or historical SEC filings.
Medical/Scientific Agents: Searching PubMed or clinical trials for verified medical information.
RAG Enrichment: Feeding high-fidelity, citation-backed data into retrieval-augmented generation pipelines.

Strengths¶

Unified API: Access to licensed repositories (PubMed, SEC, Wiley) in a single request.
Agent-Ready: Returns structured, LLM-ready data rather than just a list of links.
High Recall: Accesses "dark data" not indexable by standard search bots.
Citations: Native support for source attribution in the Answer and Deep Research endpoints.

Limitations¶

Paid Service: Requires an API key and usage-based pricing.
Latency: Searching proprietary databases can sometimes be slower than simple web-index searches.
Closed-Source: The search engine itself is a proprietary service.
Reasoning Overhead: While it provides the data, the final synthesis still depends on the reasoning capabilities of the consuming model (e.g., Claude 4.8 or GPT-5.5).

When to use it¶

When an agent needs high-accuracy, verified data from scientific, financial, or legal sources.
For building specialized agents (e.g., a "Scientific Research Agent") that require more than just web results.
To provide frontier models like claude-4-8-opus-20260528 or GPT-5.5 with grounded, verifiable context for deep reasoning tasks using high-signal research patterns.

When not to use it¶

For general, low-stakes web search where free or cheaper alternatives suffice.
If you require a fully open-source, self-hosted search index.

Getting started¶

Installation¶

Install the Valyu Python SDK via pip or uv:

pip install valyu
# or
uv add valyu

Basic Usage¶

Initialize the client and perform a simple semantic search across all sources.

from valyu import Valyu
import os

# Initialize with API key from environment
client = Valyu(api_key=os.getenv("VALYU_API_KEY"))

# Basic search query
results = client.search(query="Latest developments in room-temperature superconductors")

for result in results:
    print(f"[{result.score:.2f}] {result.title}")
    print(f"URL: {result.url}\n")

CLI examples¶

[!NOTE] Official CLI examples for Valyu are primarily managed through SDK integrations or direct API calls. A standalone CLI for end-users is not currently promoted in the official 2026 documentation.

# Example of using the Valyu API with curl
curl -X POST https://api.valyu.ai/v1/search \
  -H "Authorization: Bearer $VALYU_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"query": "Latest breakthroughs in fusion energy 2026", "source": "valyu/valyu-arxiv"}'

API examples¶

Cross-Source Answer API¶

The following example demonstrates using the Answer API to synthesize findings across scientific literature and regulatory filings.

from valyu import Valyu

# Initialize the client
client = Valyu(api_key="your-api-key")

# Perform a grounded answer query across specific proprietary sources
response = client.answer(
    query="Analyze the impact of GLP-1 agonists on healthcare provider stock volatility in 2024",
    included_sources=["valyu/valyu-pubmed", "valyu/valyu-sec-filings"],
    summary_instructions="Provide a structured analysis with citations from both medical and financial sources.",
    response_length="large"
)

print(f"Answer: {response.answer}")
for citation in response.citations:
    print(f"[{citation.id}] {citation.title} ({citation.url})")

Deep Research Pattern¶

For long-horizon tasks, use the Deep Research API to generate comprehensive reports.

from valyu import Valyu

# Initialize the client
client = Valyu(api_key="your-api-key")

# Deep Research for a specific market landscape
report = client.deep_research(
    query="Future of solid-state battery manufacturing: key players, patent landscape, and supply chain risks",
    output_format="markdown",
    max_steps=10
)

# Save the generated research report
with open("solid_state_research.md", "w") as f:
    f.write(report.content)

Sources / references¶

Contribution Metadata¶

Last reviewed: 2026-06-28
Confidence: high