Meet Hannah: The AI Presenter That Joins Your Video Calls and Runs the Deck

·1 views

A team has built an AI presenter named Hannah that joins video calls on Teams, Zoom, or Google Meet as a live participant to deliver slide decks on camera. Hannah responds to voice commands addressed to her by name, answers audience questions, and replies to chat — but only using content provided by the presenting team. To prevent misinformation, she explicitly acknowledges when a question falls outside her given material and offers to route it to a human colleague. Every session opens with a disclosure that Hannah is an AI, a step the developers say has not deterred participants and has actually encouraged more direct questioning. Key engineering challenges included managing turn-taking, knowing when to pause mid-slide, and mapping loosely worded audience questions to the correct slide content.

Read the full story at DEV Community

This is an AI-generated summary. ShortSingh links to the original source for the complete article.

Discussion (0)

EU's eIDAS 2.0 Will Make Passport-Scan KYC Legally Insufficient by 2027

The European Union's eIDAS 2.0 regulation, which entered into force in 2024, requires all EU member states to provide citizens with a government-backed European Digital Identity Wallet by end of 2026. By December 2027, large regulated entities including banks, fintechs, insurers, and telecoms must accept the EUDI Wallet for authentication. The EU's Anti-Money Laundering Regulation, on a similar 2027 timeline, further restricts identity verification to eIDAS-notified digital ID schemes, the EUDI Wallet, or qualified trust services — excluding traditional passport-scanning methods. Unlike document-based KYC, the EUDI Wallet presents cryptographically signed credentials with selective attribute disclosure, requiring platforms to validate digital signatures and trust chains rather than inspect physical ID images. KYC providers must therefore rebuild core verification pipelines to support W3C Verifiable Credentials, cryptographic trust-chain validation, and credential revocation checks ahead of the regulatory deadline.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How One Developer Built a Fast Multi-API Music Discovery Backend with FastAPI

A developer building a music app called TheExperience faced the challenge of combining data from three separate APIs — Last.fm, MusicBrainz, and the Cover Art Archive — without slowing down search performance. The solution involved splitting the backend into two distinct stages: a lightweight discovery phase triggered by user search, and a deeper enrichment phase activated only when a specific album is selected. FastAPI and asynchronous HTTP clients were used to run independent API requests concurrently, reducing latency and avoiding wasteful calls. When Last.fm provides a MusicBrainz ID for an album, the backend skips a redundant search and resolves metadata directly, cutting the number of external requests needed. This architecture treats third-party services as potentially unreliable dependencies, making the overall pipeline faster and more resilient to failures.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How to reliably trigger EAS builds via GitHub Actions using EXPO_TOKEN

Developers using Expo's EAS Build service can automate builds through GitHub Actions, but only after completing a successful local build that sets up credentials, project IDs, and build profiles. The local EAS CLI run is a mandatory prerequisite, as it creates the Android keystore, iOS certificates, and provisioning profiles that CI jobs depend on in non-interactive mode. Authentication on CI requires a personal Expo access token stored as the repository secret EXPO_TOKEN, which must be passed to the expo/expo-github-action step so all subsequent commands inherit authorization. The official workflow uses Actions checkout v5, Node 24, and the eas build --platform all --non-interactive --no-wait command to dispatch remote builds without blocking the runner. Build-time environment variables should be managed through EAS server-side profiles rather than GitHub Actions env entries, since those variables never reach the remote EAS builder.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Infere Launches AI Gateway with Prompt Versioning, Cost Tracking, and Observability

A team of developers has publicly launched Infere, an AI routing and observability platform, after months of building it to address fragmentation across multiple AI provider SDKs. The platform offers a single OpenAI-compatible endpoint that handles model routing, budget management, and logs every API call to provide clear cost visibility. It also includes git-style prompt versioning and automated evaluators to help teams manage and score prompts systematically. Infere is currently in beta and operates on a prepaid, pay-as-you-go credit model with no expiry on credits. The team is actively seeking developer feedback from real-world traffic to identify gaps and improve the platform.

0 comments Read more at DEV Community