Attachment_2_-_GenAI_Platform_Technical_Capabilities_0002.xlsx
XLSX spreadsheet 23 KB Posted
- Attached to
- NOTICE OF INTENT TO SOLE SOURCE
- Solicitation number
- DOIDFBO250048
About this file
This Excel file details technical requirements for a Generative AI (GenAI) Platform, outlining 36 comprehensive capabilities that must be implemented on Day 1 or eventually. The requirements focus on creating a user-friendly, secure, and versatile AI platform for non-technical staff, with key functionalities including a no-code chatbot interface, multi-file RAG (retrieval-augmented generation) capabilities, custom copilot creation, and robust integration options.
Critical technical specifications include Federal cybersecurity compliance (FedRAMP or AWS deployment), Web Content Accessibility Guidelines (WCAG) compliance, multi-factor authentication, Okta single sign-on, comprehensive audit logging, and cost management tools. The platform must support multiple generative AI models, allow file uploads across various formats (Microsoft Office, PDF, images, txt, csv), enable bulk analyses, provide API access, and integrate with third-party data sources while maintaining strict access permissions and data privacy standards. The opportunity is associated with the Department of the Interior Departmental Offices Interior Business Center, with a solicitation number of DOIDFBO250048, and appears to be a notice of intent to sole source.
View the file
Other files
| File | Type | Posted |
|---|---|---|
| Notice_of_Intent_to_Sole_Source_0002.pdf | ||
| Sol_DOIDFBO250048_Amd_0002.pdf | ||
| Attachment_1_-_Draft_PWS_0002.pdf | ||
| Sol_DOIDFBO250048_Amd_0001.pdf |
On GovTribe
Work with this file
- Download the original file
- Contacts named in this file
- Similar government files
- Ask GovTribe AI about this file
Text version
GenAI Platform Capabilities
| Req ID | Must Have Requirements | ||
| "Users" refers to non-technical staff, not specialized data or AI analysts, engineers, etc. Availability should describe current-state functionality. | Importance | Vendor Name (Yes/No) | |
| 1 | Users can engage with genAI models through chatbot interface with no code. | Must Have on Day 1 | |
| 2 | Users can see and navigate through their history of chats. | Must Have on Day 1 | |
| 3 | Users can rename chats from auto-generated chat names. | Must Have on Day 1 | |
| 4 | Users can include file attachments with prompts, which are then referenced through full context or, if exceeding context window, retrieval-augmented generation (RAG). | Must Have on Day 1 | |
| 5 | Users can manually upload Microsoft office files (Word, Excel, PowerPoint) for RAG | Must Have on Day 1 | |
| 6 | Users can manually upload PDF files for RAG | Must Have on Day 1 | |
| 7 | Users can manually upload image files for RAG | Must Have on Day 1 | |
| 8 | Users can manually upload .txt files for RAG | Must Have on Day 1 | |
| 9 | Users can manually upload .csv files for RAG | Must Have on Day 1 | |
| 10 | Users can specify specific web pages by URL to include as data sources for RAG. Users have option to also include the sub-pages of the specified URLs. | Must Have on Day 1 | |
| 11 | Users can organize and curate collections of data sources that can be used for multiple purposes. | Must Have on Day 1 | |
| 12 | Users can create "copilots", i.e. specialized virtual assistants that have stored context and established personas and prompts. Users should be able to create copilots without code. | Must Have on Day 1 | |
| 13 | Users can test and evaluate copilot responses within the genAI platform. | Must Have on Day 1 | |
| 14 | Users can deploy copilots in a variety of environments including the web interface and API-integrated systems. | Must Have on Day 1 | |
| 15 | Users can provide feedback to copilots, which should then adapt its responses and actions iteratively based on that feedback. | Must Have on Day 1 | |
| 16 | Users can configure copilots with agentic capabilities including but are not limited to: calling other copilots; accessing and using API services; and executing automated workflows. | Must Have on Day 1 | |
| 17 | Users can conduct bulk analyses without code. Bulk analyses are defined as making the same series of large language model calls to extract information into a structured format from a collection of documents or spreadsheet. | Must Have on Day 1 | |
| 18 | GenAI platform allows users to select between different generative AI models provided by multiple developers | Must Have on Day 1 | |
| 19 | GenAI platform can support large language models. | Must Have on Day 1 | |
| 20 | GenAI platform meets Federal cybersecurity and privacy standards applicable to systems handling personally identifiable information (PII) and protected health information (PHI) at a moderate risk level. Software-as-a-service platforms must be authorized by FedRAMP. Alternatively, the platform must support deployment in a government-owned Amazon Web Services (AWS) environment to maintain control over sensitive data. Compliance measures must be clearly documented, including encryption, access controls, and auditing tools, to aid government agencies in meeting mandated security and privacy requirements. | Must Have on Day 1 | |
| 21 | GenAI platform meets the Web Content Accessibility Guidelines (WCAG) Level 2.1 AA requirements so that it is compliant with Section 508 of the Rehabilitation Act. | Must Have on Day 1 | |
| 22 | GenAI platform is designed with a user-friendly interface that prioritizes simplicity and intuitive navigation for system administrators and users of all skill levels. GenAI platform user interface includes clear instructions, accessible tutorials, and contextual guidance to ensure users can easily understand and utilize its features. The platform should minimize technical jargon and provide seamless workflows and customization options, enabling users to achieve their desired outputs with minimal effort. | Must Have on Day 1 | |
| 23 | GenAI platform has responsive customer support. | Must Have on Day 1 | |
| 24 | GenAI platform has robust error handling mechanisms. | Must Have on Day 1 | |
| 25 | GenAI platform can directly integrate with third-party data sources including SaaS products and self-hosted. | Must Have on Day 1 | |
| 26 | GenAI platform honors all access permissions set in all integrated data source systems. | Must Have on Day 1 | |
| 27 | GenAI platform has well-documented APIs that allow developers to access and interact with all platform features programmatically. | Must Have on Day 1 | |
| 28 | GenAI platform refreshes integrated data so that users have the most up-to-date information when using data from an integrated data source. | Must have eventually | |
| 29 | Users must use multi-factor authentication | Must have on Day 1 | |
| 30 | Users can authenticate into genAI platform through Okta single sign-on. | Must have eventually | |
| 31 | GenAI platform includes redaction policies that allow user and system prompts to be scrubbed for specified types of information (e.g. phone numbers) or key phrases. | Must Have on Day 1 | |
| 32 | GenAI platform has audit logs that allow the genAI platform administrators to see the history of all prompts and responses by user, copilot, model, timestamp, and API key. | Must Have on Day 1 | |
| 33 | GenAI platform audit logs provide visibility into cost and performance metrics (e.g. completion time) for each interaction. | Must have eventually | |
| 34 | GenAI platform audit logs are exportable in CSV format and accessible via API. | Must Have on Day 1 | |
| 35 | GenAI platform has cost management tools that allow the genAI platform administrators to set maximum cost thresholds at multiple granularities. | Must have eventually | |
| 36 | GenAI platform has a clear return-on-investment and a sustainable cost-per-user-per-month, inclusive of the continuously developed platform software, cloud hosting costs, and genAI model costs. | Must Have on Day 1 | |
| Additional | Provide Pricing Model and Tiered Discount Information |
File details come from the government source that posted it. Updated .