| a2dd2c933292cefd… | For ingest, transform, and refresh, the Fabric-native column identifies Fabric pipelines plus Purview as the mechanism used to manage those data operations. | active | contextual | yes | |
| 36105045c16d294f… | For the requirement of where data is stored, the Fabric-native approach uses governed OneLake, while the recommended Pyrana-on-Fabric approach leaves OneLake as the system of record and stores only cited claims and provenance in Pyrana’s knowledge graph within the customer’s Azure tenant. | active | contextual | yes | |
| 3d9a13dcc1197557… | The Platform and IT visibility RFP requirement states that IT must have visibility and approval over data storage, ingestion, transformation, refresh, security, access, sensitive-data protection, production deployment changes, and the governance, testing, and monitoring of AI agents. | active | contextual | yes | |
| e007d13d89757b0f… | Section 5 presents a requirement-by-requirement crosswalk that places every Summer Fridays RFP requirement category beside both the Fabric-native approach and the Pyrana-on-Fabric approach so that the response can serve as the backbone of a compliant, differentiated vendor proposal. | active | contextual | yes | |
| ff415c60b8c10d09… | The document frames the key distinction between Pyrana and a Fabric plus Foundry custom build as a build-versus-buy decision centered on governance ownership and delivery time, which it says is addressed in Section 6. | active | contextual | yes | |
| 76303d631132ff53… | The middle-column interpretation is that Fabric plus Foundry plus custom build can reach the same end state as Pyrana because Microsoft supplies capable building blocks, even though the governance layer must still be built, hardened, and maintained by the buyer. | active | contextual | yes | |
| d1e5072e8cdf52fe… | The document states that Fabric Data Agent limitations in unstructured grounding and observability reflect Microsoft’s documented 2026 capabilities and are expected to continue evolving over time. | active | contextual | yes | |
| 4888676a27e04ed6… | The comparison legend defines a check mark as provided or strong, a triangle as partial or build-it-yourself, and a cross as not provided so that the capability matrix can be interpreted consistently. | active | contextual | yes | |
| 4f1df5b6045aa7e3… | Time to production for governed agents is fast but thin with Fabric Data Agents, longest with a Fabric plus Foundry custom build because the user builds it, and reduced to weeks with Pyrana because it is pre-governed. | active | contextual | yes | |
| 30bd86aaac9d7c73… | All three compared approaches stay in Azure and honor the Fabric mandate, with Pyrana specifically described as Azure-native and able to read OneLake. | active | contextual | yes | |
| 73c1793da2af12f5… | Custom business-facing agent user interfaces are not provided by Fabric Data Agents because Power BI is positioned as BI, require building a custom app in a Fabric plus Foundry custom build, and are accelerated in Pyrana through UI kits and an SDK that can deliver in weeks. | active | contextual | yes | |
| 2fbf320a1722a0eb… | Prompt-injection and misuse protection are only platform-level in Fabric Data Agents, must be implemented by the user in a Fabric plus Foundry custom build, and are built into governance in Pyrana on Fabric. | active | contextual | yes | |
| 489dff52c2f66468… | Agent-level observability and evaluation are a current gap in Fabric Data Agents, require enforced conventions and assembled tooling in a Fabric plus Foundry custom build, and are provided in Pyrana through OTel, Langfuse, and MLflow. | active | contextual | yes | |
| 8da0c806d2a7fd6a… | Deterministic claim checks under the principle 'Validate, don't invent' are not provided by Fabric Data Agents, are absent in the Fabric plus Foundry custom build column, and are implemented in Pyrana through a generate-then-validate pattern. | active | contextual | yes | |
| 53be87a48ea7a6ff… | Fact-versus-recommendation-versus-assumption typing is not provided by Fabric Data Agents, requires custom guardrails in a Fabric plus Foundry custom build, and is supported in Pyrana through knowledge-type labels and confidence values. | active | contextual | yes | |
| 0a8ac2203e47977f… | Source-to-AI citation on every answer is only partial in Fabric Data Agents, must be designed manually in a Fabric plus Foundry custom build, and is supported in Pyrana by logging Context Unit IDs. | active | contextual | yes | |
| 5ce6d080ca755ed2… | Human-in-the-loop approval gates are not available in Fabric Data Agents, require custom implementation in a Fabric plus Foundry custom build, and are provided natively in Pyrana on Fabric. | active | contextual | yes | |
| 4828cd86b319629a… | Durable multi-step, multi-agent workflows are not provided by Fabric Data Agents, require user orchestration through Agent Framework in a Fabric plus Foundry custom build, and are provided by Pyrana through Harness plus Temporal. | active | contextual | yes | |
| f6fbcca8cf846041… | Grounding on unstructured documents such as PDFs, Word files, and decks is not native in Fabric Data Agents, is only partial in a Fabric plus Foundry custom build because retrieval must be built manually, and is provided by Pyrana through its Context Engine. | active | contextual | yes | |
| bf66f213b704db5e… | Fabric Data Agents, Fabric plus Foundry plus custom build, and Pyrana on Fabric all support governed question answering over structured data, with Pyrana specifically reading the Fabric model so that structured-data access is available in every compared approach. | active | contextual | yes | |
| 2e6cc4e9ab70f279… | The head-to-head comparison matrix evaluates the agentic-application capabilities required by the project, while Fabric’s data and BI strengths are treated as already stipulated and covered separately in Section 7. | active | contextual | yes | |
| 5b48e59e8f4ddb07… | The architecture applies a security rail across all layers using Entra ID identity, Azure Key Vault, AES-256 encryption at rest, TLS 1.2 or higher in transit, private networking, and per-tenant isolation so that it remains consistent with Pyrana’s Azure production architecture. | active | contextual | yes | |
| 8e92ecbf7b1bde00… | The unstructured knowledge path uses Pyrana Context Engine to extract atomic, cited Context Units from recap documents, calendars, claim libraries, and PDFs so that agents can retrieve and cite grounded knowledge coverage that Fabric Data Agents do not natively provide. | active | contextual | yes | |
| 059710789f61c064… | The structured data path routes source systems through Fabric pipelines into OneLake or a warehouse and then into a governed semantic model so that Power BI dashboards and Pyrana agents can query the same governed source of truth. | active | contextual | yes | |
| 8547dec6a1d5e548… | The head-to-head capability comparison is intended to score the agentic-application capabilities required by the project, while Fabric's data and BI strengths are treated as already stipulated and covered elsewhere. | active | contextual | yes | |
| 2aa257e5b51de589… | A security rail must apply across all layers using Entra ID identity, Azure Key Vault, AES-256 encryption at rest, TLS 1.2 or higher in transit, private networking, and per-tenant isolation so that the architecture matches Pyrana's Azure production deployment. | active | contextual | yes | |
| 0bcae69cd3c268f6… | For the unstructured knowledge path, recap documents, launch and marketing calendars, claim libraries, and brand-health and competitive PDFs must be processed by the Pyrana Context Engine to extract atomic, cited Context Units so that agents can retrieve and cite them. | active | contextual | yes | |
| 3411bd97721e7081… | For the structured data path, sources must flow through Fabric pipelines into OneLake or a warehouse and then into a governed semantic model so that Power BI dashboards and Pyrana agents query the same governed model and reconcile to one source of truth. | active | contextual | yes | |
| 59a4b2e016911668… | The architecture ingests business data from commerce, marketplace, media, influencer, sentiment, competitive, planning, budget, and recap sources so that both structured and unstructured workflows can operate on relevant business inputs. | active | contextual | yes | |
| 4ce402f06170fdad… | The data foundation layer consists of OneLake, Lakehouse or Warehouse storage, pipelines, semantic models, and Purview governance and lineage, with optional Fabric Data Agents for Q&A so that governed data services remain in Fabric. | active | contextual | yes | |
| 5d94b54f861a7955… | The orchestration layer uses PyranaHarness with Temporal for durable execution and HITL, Cortex Context Units for cited and versioned context, a tiered code interpreter, OTel or Langfuse telemetry, and audit plus approvals so that agent operations are governed and traceable. | active | contextual | yes | |
| b91e29fc263369e4… | The presentation layer consists of Power BI for BI dashboards, a Pyrana custom frontend for agent user experience and HITL, and Entra ID single sign-on so that users access both analytics and agent workflows through governed identity. | active | contextual | yes | |
| 3cf7b6c2d5ebe16f… | The business-facing experience should provide governed dashboards and an AI-agent workspace with human-in-the-loop review for teams including Marketing, eCommerce, Retail, Media, and Brand Health so that users can access analytics and reviewed agent outputs. | active | contextual | yes | |
| d69163d5fa3c9b33… | An exception is considered reasonable because the RFP expects exceptions to be documented and justified, and using Pyrana avoids the greater scope, longer timeline, and governance burden of hand-building agent controls inside Fabric. | active | contextual | yes | |
| e9a49454666eb316… | Fabric must remain the data foundation, dashboard surface, and governance authority of record while Pyrana supplies supporting architecture, data-flow, and security documentation to justify approval of the exception. | active | contextual | yes | |
| 531380faa200dae5… | The written exception rationale states that Pyrana delivers AI-agent capability as an Azure-native agentic application deployed in the tenant and consuming governed data from OneLake without writing data outside Azure so that the exception remains bounded within Azure. | active | contextual | yes | |
| 4d46c004248a7eac… | IT should document and approve Pyrana as the single platform rather than allowing a sprawl of external tools so that governance and oversight remain centralized. | active | contextual | yes | |
| 810be5bab541637b… | The platform must use private inference with per-tenant isolation and must not train models on Summer Fridays data so that customer data remains isolated and is not reused for model training. | active | contextual | yes | |
| 8e737a67e5950854… | Pyrana must read only from governed Fabric data, and raw confidential or customer-level data must never be exposed through dashboards or agents unless explicit approval is granted so that sensitive data access is controlled. | active | contextual | yes | |
| f9d7656f7891ca04… | Pyrana must run inside Summer Fridays' Azure tenant using Azure OpenAI or Foundry models, Azure Postgres, Entra ID, and AKS so that the agent platform remains inside the Microsoft estate. | active | contextual | yes | |
| 6bddfa40bcbe8536… | Microsoft Purview must remain the system that owns cataloging, lineage, sensitivity labels, and DLP controls so that governance authority stays with Fabric-native governance services. | active | contextual | yes | |
| 570d73d445a4bd07… | Dashboards must reconcile to approved Microsoft Fabric logic and Power BI reports so that business-facing analytics remain aligned with the sanctioned data model. | active | contextual | yes | |
| 0f2156768d49f125… | The head-to-head capability comparison is scoped to the agentic-application capabilities required by the project, while Fabric’s data and BI strengths are treated as already stipulated and covered elsewhere. | active | contextual | yes | |
| 3e19fb62f7b533af… | A security rail applies across all architecture layers using Entra ID identity, Azure Key Vault, AES-256 encryption at rest, TLS 1.2 or higher in transit, private networking, and per-tenant isolation consistent with Pyrana’s Azure production architecture. | active | contextual | yes | |
| e7136ab631f70134… | Context Units are defined here as atomic, cited units extracted by the Pyrana Context Engine so that agents can retrieve and cite discrete pieces of unstructured knowledge. | active | contextual | yes | |
| ec821c0b562c680f… | The knowledge path for unstructured content sends recap documents, launch and marketing calendars, claim libraries, and brand-health or competitive PDFs into the Pyrana Context Engine, which extracts atomic cited Context Units for agent retrieval and citation. | active | contextual | yes | |
| 4f069f6686405167… | The structured data path runs from sources through Fabric pipelines into OneLake or a warehouse and then into a governed semantic model so that Power BI dashboards and Pyrana agents query the same governed model and reconcile to one source of truth. | active | contextual | yes | |
| 0fa6b38574c33ba5… | The architecture draws from commerce, media, and brand information sources including Shopify, Amazon, Sephora portals, TikTok Shop, affiliates, paid media, influencer and EMV data, reviews and sentiment, competitive intelligence, calendars, budgets, and recap documents. | active | contextual | yes | |
| 741b0f1149aa0016… | The orchestration and governance layer includes PyranaHarness with Temporal for durable HITL workflows, Cortex Context Units for cited and versioned context, a tiered code interpreter, OTel or Langfuse telemetry, and audit plus approvals. | active | contextual | yes | |
| 0fb52febd9b8e78f… | The presentation layer consists of Power BI for BI dashboards, a Pyrana custom frontend for agent user experience and HITL, and Entra ID single sign-on for access control. | active | contextual | yes | |