API Documentation
Objective is a set of plug-and-play APIs that make it easy to ship & iterate on high quality AI-powered search. Get up and running in less than 5 minutes, and get amazing results out of the box.
Getting started
To get started, create a new Objective account, then read about how to make requests for the resources you need to access using our HTTP APIs or dedicated client SDKs.
Demo (3 mins)
Features
- Indexing
- Text, image & multimodal indexes - Index and search semantically on your text and image data using state of the art vector embeddings.
- Hybrid Indexes (Default) - Combine lexical (keyword) and semantic (neural) search capabilities to find both keyword and semantic matches, ensuring users receive the most relevant results for precise terms or broader concepts.
- Real-time indexing
- Document processing & image crawling
- Schemaless - Index both structured or unstructured data.
- High throughput indexing - Index 1000s of objects per second.
- Evaluate and improve relevance
- Evaluations - Run evaluations and compare the search quality of your indexes and track the quality over time.
- Side by side tool - Side by side evaluations and search results to compare search systems.
- Finetuning - Train your indexes to improve relevance based on user feedback or your specific business needs.
- Auto-finetuning - Train your indexes using state of the art AI models when human feedback is not readily available.
- Result Cutoffs - Dynamically limit the number of results to only the most relevant.
- Search
- Semantic similarity - Return content-based recommendations to users.
- Hybrid Search - Leverages both keyword (lexical) and semantic (neural) signals to deliver accurate and comprehensive search results.
- Filtering - Extensive and flexible syntax for dynamically filtering search results.
- Sorting - Sort and rank results.
- Geo-filtering - Filter search results based on geographic location.
- Ranking Signals - Dynamically change or personalize search results at query-time using a robust and extensive syntax.
- Ranking Refinement - Automatically filter search results to only the most relevant.
- Highlights - Highlight specific elements in your search results based on their relevance to the user's query.
- Low latency - Average 40ms search request latency.
- Platform
- Highly scalable infra - horizontally and elastically scalable, low latency, high reliability.
- REST APIs, Python and Typescript SDKs.
- Developer friendly - Go from zero to production grade search API in minutes.
- Integrations with frameworks such as Langchain.
Platform overview
At its core, the Objective platform stores your Objects and enables you to build Indexes from those Objects for the purpose of searching.
Objects are schemaless, and can contain both structured and unstructured data. You'll use the Objects API to store and manage Objects in the Object Store.
Objects in the Object Store can then be added to one or more Indexes using the Create Index API. Being added to an Index makes an Object searchable. When creating an Index, you're able to choose specific fields from your Objects to index, such as title
, description
and image_url
. Indexes are kept up to date as you make updates to the underlying Objects in the Object Store.
Finally, after having added your Objects to one or more Indexes, you'll use the Search API to issue search requests against your Indexes.
Objective Search API integration model
download as pdf
Key concepts
The Object Store and Objects
All data added to the platform is stored in the Object Store. The Object Store is a collection of schemaless Objects, which are represented using JSON. An Object's fields can be searchable text (represented as Strings), URLs to images that the platform should crawl (also represented as Strings) as well as other JSON data types (arrays, numbers, dates, booleans, and nested JSON).
Indexes
Indexes are built from Objects in the Object Store, according to composition rules that you define. Adding Objects to an Index is what makes them searchable. Once an Index is built, you can query the Objects in that Index using the /search
API.
Highlights
When you search an Index, the API returns relevant Objects along with Highlights. A Highlight represents the most relevant portions of a matching Object, be it a section of text or an image.