Navigating a sea of paperwork, scattered throughout varied platforms, generally is a daunting process, usually resulting in gradual decision-making and missed insights. As organizational data and knowledge multiplies, groups that may’t centralize or floor the fitting info shortly will battle to make choices, innovate, and keep aggressive.
This weblog explores how the brand new Talk to My Docs (TTMDocs) agent gives an answer to the steep prices of data fragmentation.
The excessive value of data fragmentation
Information fragmentation is not only an inconvenience — it’s a hidden value to productiveness, actively robbing your group of time and perception.
- A survey by Starmind throughout 1,000+ data employees discovered that staff solely faucet into 38% of their obtainable data/experience as a result of of this fragmentation.
- One other study by McKinsey & Associates found that data employees spend over 1 / 4 of their time looking for the knowledge they want throughout completely different platforms corresponding to Google Drive, Field, or native techniques.
The constraints of current options
Whereas there are a number of choices in the marketplace designed to ease the method of querying throughout key paperwork and supplies residing in quite a lot of locations, many have vital constraints in what they will really ship.
For instance:
- Vendor lock-in can severely hinder the promised expertise. Until you’re strictly utilizing the supported integrations of your vendor of selection, which in most cases is unrealistic, you find yourself with a restricted subset of data repositories you may connect with and work together with.
- Safety and compliance concerns add one other layer of complexity. If in case you have entry to at least one platform or paperwork, it’s possible you’ll not want entry to a different, and any misstep or missed vulnerability can open up your group to potential threat.
Speak to My Docs takes a unique strategy
DataRobot’s new Talk to My Docs agent represents a unique strategy. We offer the developer instruments and assist you should construct AI options that really work in enterprise contexts. Not as a vendor-controlled service, however as a customizable open-source template you may tailor to your wants.
The differentiation isn’t refined. With TTMDocs you get:
- Enterprise safety and compliance in-built from day one
- Multi-source connectivity as a substitute of vendor lock-in
- Zero-trust entry management (Respects Current Permissions)
- Full observability by way of DataRobot platform integration
- Multi-agent structure that scales with complexity
- Full code entry and customizability as a substitute of black field APIs
- Fashionable infrastructure-as-code for repeatable deployments
What makes Speak to My Docs completely different
Speak To My Docs is an open-source utility template that provides you the intuitive, acquainted chat-style expertise that trendy data employees have come to count on, coupled with the management and customizability you really need.
This isn’t a SaaS product you subscribe to; however reasonably a developer-friendly template you may deploy, modify, and make your personal.
Multi-source integration and actual safety
TTMDocs connects to Google Drive, Field, and your native filesystems out of the field, with Sharepoint and JIRA integrations coming quickly.
- Protect current controls: We offer out-of-the-box OAuth integration to deal with authentication securely by way of current credentials. You’re not making a parallel permission construction to handle—if you happen to don’t have permission to see a doc in Google Drive, you gained’t see it in TTMDocs both.
- Meet knowledge the place it lives: Not like vendor-locked options, you’re not compelled emigrate your doc ecosystem. You’ll be able to seamlessly leverage recordsdata saved in structured and unstructured connectors like Google Drive, Field, Confluence, Sharepoint obtainable on the DataRobot platform or add your recordsdata regionally.
Multi-agent structure that scales
TTMDocs makes use of CrewAI for multi-agent orchestration, so you may have specialised brokers dealing with completely different elements of a question.
- Modular & versatile: The modular structure means you too can swap in your most well-liked agentic framework, whether or not that’s LangGraph, LlamaIndex, or every other, if it higher fits your wants.
- Customizable: Wish to change how brokers interpret queries? Modify the prompts. Want customized instruments for domain-specific duties? Add them. Have compliance necessities? Construct these guardrails instantly into the code.
- Scalable: As your doc assortment grows and use circumstances change into extra complicated, you may add brokers with specialised instruments and prompts reasonably than attempting to make one agent do all the pieces. For instance, one agent would possibly retrieve monetary paperwork, one other deal with technical specs, and a 3rd synthesize cross-functional insights.
Enterprise platform integration
One other key side of Speak to my Docs is that it integrates along with your current DataRobot infrastructure.
- Guarded RAG & LLM entry: The template features a Guarded RAG LLM Mannequin for managed doc retrieval and LLM Gateway integration for entry to 80+ open and closed-source LLMs.
- Full observability: Each question is logged. Each retrieval is tracked. Each error is captured. This implies you’ve gotten full tracing and observability by way of the DataRobot platform, permitting you to truly troubleshoot when one thing goes flawed.
Fashionable, modular elements
The template is organized into clear, unbiased items that may be developed and deployed individually or as a part of the complete stack:
| Element | Description |
| agent_retrieval_agent | Multi-agent orchestration utilizing CrewAI. Core agent logic and question routing. |
|
core |
Shared Python logic, widespread utilities, and features. |
| frontend_web | React and Vite internet frontend for the consumer interface. |
| internet | FastAPI backend. Manages API endpoints, authentication, and communication. |
| infra | Pulumi infrastructure-as-code for provisioning cloud assets. |
The ability of specialization: Speak to My Docs use circumstances
The sample is productionized specialised brokers, working collectively throughout your current doc sources, with safety and observability in-built.
Listed below are a number of examples of how that is utilized within the enterprise:
- M&A due diligence: Cross-reference monetary statements (Field), authorized contracts (Google Drive), and technical documentation (native recordsdata). The permission construction ensures solely the deal group sees delicate supplies.
- Scientific trial documentation: Confirm trial protocols align with regulatory pointers throughout a whole bunch of paperwork, flagging inconsistencies earlier than submission.
- Authorized discovery: Search throughout years of emails, contracts, and memos scattered throughout platforms, figuring out related and privileged supplies whereas respecting strict entry controls.
- Product launch readiness: Confirm advertising supplies, regulatory approvals, and provide chain documentation are aligned throughout areas and backed by certifications.
- Insurance coverage claims investigation: Pull coverage paperwork, adjuster notes, and third-party assessments to cross-reference protection phrases and flag potential fraud indicators.
- Analysis grant compliance: Cross-reference finances paperwork, buy orders, and grant agreements to flag potential compliance points earlier than audits.
Use case: Scientific trial documentation
The problem
A biotech firm getting ready an FDA submission is drowning in documentation unfold throughout a number of techniques: FDA steerage in Google Drive, trial protocols in SharePoint, lab studies in Field, and high quality procedures regionally. The core downside is making certain consistency throughout all paperwork (protocols, security, high quality) earlier than a submission or inspection, which calls for a fast, unified view.
How TTMDocs helps
The corporate deploys a personalized healthcare regulatory agent, a unified system that may reply complicated compliance questions throughout all doc sources.
Regulatory agent:
Identifies relevant FDA submission necessities for the precise drug candidate.
Scientific evaluation agent:
Evaluations trial protocols towards trade requirements for affected person security and analysis ethics.
Security compliance agent:
Checks that security monitoring and hostile occasion reporting procedures meet FDA timelines.
The end result
A regulatory group member asks: “What do we want for our submission, and are our security monitoring procedures as much as normal?”
As a substitute of spending days gathering paperwork and cross-referencing necessities, they get a structured response inside minutes. The system identifies their submission pathway, flags three high-priority gaps of their security procedures, notes two points with their high quality documentation, and gives a prioritized motion plan with particular timelines.
The place to look: The code that makes it occur
One of the best ways to know TTMDocs is to have a look at the precise code. The repository is totally open supply and obtainable on Github.
Listed below are the important thing locations to start out exploring:
- Agent structure (agent_retrieval_agent/custom_model/agent.py): See how CrewAI coordinates completely different brokers, how prompts are structured, and the place you may inject customized habits.
- Instrument integration (agent_retrieval_agent/custom_model/device.py): Reveals how brokers work together with exterior techniques. That is the place you’d add customized instruments for querying an inner API or processing domain-specific file codecs.
- OAuth and safety (internet/app/auth/oauth.py): See precisely how authentication works with Google Drive and Field and the way your consumer permissions are preserved all through the system.
- Net backend (internet/app/): The FastAPI utility that ties all the pieces collectively. You’ll see how the frontend communicates with brokers, and the way conversations are managed.
The way forward for enterprise AI is open
Enterprise AI is at an inflection level. The hole between what end-user AI instruments can do and what enterprises really need is rising. Your organization is realizing that “ok” client AI merchandise create extra issues than they resolve once you can not compromise on enterprise necessities like safety, compliance, and integration.
The long run isn’t about selecting between comfort and management. It’s about having each. Talk to my Docs places each the ability and the pliability into your fingers, delivering outcomes you may belief.
The code is yours. The probabilities are infinite.
Expertise the distinction. Begin constructing right this moment.
With DataRobot utility templates, you’re by no means locked into inflexible black-box techniques. Acquire a versatile basis that permits you to adapt, experiment, and innovate in your phrases. Whether or not refining current workflows or creating new AI-powered purposes, DataRobot provides you the readability and confidence to maneuver ahead.
Begin exploring what’s doable with a free 14-day trial.
