What is GroundCite?
GroundCite is an advanced Gemini-based multi-agent library designed to improve research accuracy and reliability. It enhances Gemini with source filtering, citation validation, and structured output formatting through a multi-stage pipeline. This ensures comprehensive, verifiable, and well-grounded research results.
Why is GroundCite important for Research?
-
Prevents broken citations that hurt trust.
-
Improves Google E-E-A-T (Experience, Expertise, Authoritativeness, Trustworthiness) signals.
-
Provides structured JSON outputs for integration with RAG, data pipelines, and knowledge graphs.
-
Ensures compliance-ready reporting for regulated industries.
Who should use GroundCite?
GroundCite is ideal for compliance ready reporting for regulated industries like esp industries like Healthcare, Law and Research Firms etc where information credibility matters, including:
-
– Ensure every article is backed by reliable sources.
-
– Avoid broken or irrelevant citations in scholarly work.
-
– Deliver trustworthy insights with verifiable references.
If you’re using Gemini for research, GroundCite helps eliminate unreliable outputs and adds layers of validation and grounding with structured, citation-rich results.
What features does GroundCite add to Gemini?
GroundCite extends Gemini’s capabilities with:
-
Configurable Source Preferences – Prioritize trusted domains like
.gov,.edu, or official sites. -
Domain Blacklisting – Exclude unwanted or unreliable sites permanently.
-
Citation Validation – Ensure every source is live, accessible, and relevant.
-
Relevance Validation – Filter out “next best” irrelevant results.
-
Inline Numbered Citations – Every fact maps to a verifiable source.
-
JSON Outputs with Citations – Enriched text, then parsed into structured format.
-
Configurable Strategies – Choose parsing models, set strictness levels, and manage confidence thresholds.
Does GroundCite store my Gemini or OpenAI API keys?
No, your keys are never stored. GroundCite operates locally or within your secure environment.
How to install GroundCite?
Install via pip:
Which API keys are required?
-
Mandatory: Google Gemini API key
-
Optional: OpenAI API key – for structured JSON outputs with citations.
How to run a basic query with GroundCite?
Example:
How do I enable validation?
How does the validation logic work?
GroundCite uses two layers of validation:
-
Citation Validation – Checks if the source is live, accessible, and relevant.
-
Relevance Validation – Uses AI to ensure content truly matches the reference, with a confidence threshold of 0.7 or higher.
Only verified content is retained for maximum trustworthiness.
How do I extract structured data (parsing)?
How do I filter search results by specific sites?
Include only selected sites:
Exclude unwanted sites:
Does --include-sites fetch only from included sites?
Yes. When you use --include-sites, GroundCite restricts the content exclusively to the specified domains, rather than just prioritizing them.
How do I use the GroundCite REST API?
Start the server:
Then send a POST request to:
with your query payload.
Where can I find out more about GroundCite: Code, Architecture , usage etc?
Please visit out github repo at https://github.com/cennest/ground-cite