The SeekrFlow Agents API is a powerful tool for building agentic AI applications. It can be used to design autonomous agents with access to advanced tools and models—unlocking your ability to develop sophisticated virtual advisors, personalized learning companions, predictive maintenance systems, and intelligent workflow automations tailored to your unique needs.

What are agents?

Agents are designed to operate independently: perceiving their environment, making decisions, and taking actions based on the goals set by you, the user. Unlike simple chatbots or search engines, which primarily respond to user input, agents can recall previous interactions, adapt their approach based on available information, and perform complex tasks that might otherwise require human intervention.

Agents are characterized by their:

Environment: This is defined by the use case; e.g., a CRM system for sales analytics, or a corporate knowledge base for employee onboarding.

Available toolkit: The agent's capabilities expand based on the specific tools and functions it can access.

Tools are what allow agents to interact with the outside world. They fall into three categories:

1. Knowledge augmentation: Context construction, Internet search, vector search, SQL queries.
2. Capability extension: Calculators, code interpreters, computer-using agents that interact with digital environments like a human.
3. Action-taking: Chatbots that can perform tasks (resetting passwords, processing refunds); autonomous vehicles. These can set their own goals and interact with the physical world, which makes them the most powerful (and potentially most dangerous) of these categories.

Available tools

Currently, the Agents API supports the following tools:

• FileSearch: A knowledge augmentation tool that processes and searches files.
• WebSearch: A knowledge augmentation tool that searches across multiple web sources and search engines for real-time information.
• Code Interpreter: This tool enables agents to execute Python code in a secure environment, making it ideal for data analysis, calculations, file processing, and generating visualizations.

The SeekrFlow agent advantage

Seekr's API delivers a fully hosted infrastructure and unified framework, letting you fine-tune, deploy, and run advanced models—all within your own secure environment. Unlike other frameworks, which require stitching together multiple third-party services and external dependencies, Seekr keeps your data and completions ringfenced, supporting air-gapped and on-premise deployments for maximum control and compliance.

Users have the ability to create and plug in custom fine-tuned or RAG models within our agentic framework, using dedicated endpoints. Seekr is purpose-built for extensibility and enterprise needs, giving you a robust, secure base to grow with.

API-powered context management and continuity

Our API uses threads and messages for context management and efficiency. Threads are containers that automatically store conversation history in the form of messages, or individual interactions.

Example message:

"messages": [ 

    {"role": "user", "content": "How do I reset my password?"}, 

    {"role": "assistant", "content": "To reset your password, go to the account settings and click on 'Reset Password'."} 

] 

Threads persist across sessions, allowing you to return to ongoing discussions without losing context. Threads are scalable to large numbers of users, while retaining context for individual users.

Vector databases for document structuring and knowledge retrieval

Vector databases power fast, accurate semantic search by storing and indexing high-dimensional vector embeddings, which capture the meaning and context of files, rather than just keywords.

After you create a vector database and add documents to it, a four-step process is initiated:

1. Conversion: Documents are converted into plain text
2. Chunking: The text is split into smaller, semantically meaningful segments (512-1024 tokens)
3. Embedding: Each segment is converted into high-dimensional vector embeddings
4. Indexing: Embeddings are organized into an index for high-efficiency similarity searches

Web search capabilities for real-time information

The WebSearch tool enables agents to access current information from across the web, complementing the knowledge stored in vector databases. This tool provides:

Search Execution

• Natural language processing: Agents can submit text queries in natural language directly to the search tool
• Performance: Returns search results within reasonable time limits based on search complexity
• Error handling: Failed searches provide clear error messages to help agents refine their approach

Search Scope and Coverage

• Multi-source access: Searches across multiple web sources and comprehensive search engines
• Diverse content types: Returns results from various content types including news, blogs, academic papers, and government publications
• Real-time content: Can discover content published within the last 24 hours for up-to-date information
• Global reach: Supports international and multilingual content discovery

Context-aware integration

• Conversation context: Uses conversation history to improve search relevance and accuracy
• Follow-up understanding: Previous search results inform and enhance subsequent searches
• User preference learning: Search history improves results over time

Code interpreter capabilities for data analysis and computation

The code interpreter tool gives agents the ability to run Python code dynamically, unlocking advanced data analysis, mathematical computation, and visualization capabilities. This tool provides:

Code execution

• Secure environment: Python runs in a sandboxed environment with no data persistence beyond execution
• Package support: Common data science libraries (pandas, numpy, matplotlib) are readily available
• File processing: Can analyze uploaded files, datasets, and documents
• Results integration: Code execution results are returned to agents for use in workflows

Analysis and computation

• Data analysis: Process CSVs, perform statistical calculations, and extract insights from datasets
• Visualizations: Create charts, graphs, and interactive visualizations using matplotlib and seaborn
• Mathematical operations: Execute complex calculations and mathematical computations
• Report generation: Generate data-driven reports and summaries with quantitative analysis