SShortSingh.
TopTechnologyAI & MLStartupsProgrammingScienceBusinessFinanceCrypto & Web3World
Back to feed
Programming

Dev team builds AI agent to convert messy RFQ emails into priced quotes in seconds

0
·2 views

A development team spent three weeks building Distill.ai, an AI-powered tool that automates the process of turning unstructured request-for-quote emails and PDFs into accurate sales quotes. The system runs a seven-stage pipeline covering parsing, extraction, classification, catalog matching, pricing, policy checks, and confidence scoring. Any line item scoring below a 0.70 match threshold is automatically flagged and routed to a human reviewer rather than quoted, a design choice the team says makes the tool practically trustworthy for sales teams. The tool was built using Alibaba Cloud Model Studio, leveraging Qwen-Plus for extraction and classification and text-embedding-v4 for vector-based catalog matching via pgvector in Postgres. The application was deployed in Singapore to minimize latency with the Model Studio endpoint, and runs on a NestJS TypeScript backend with Redis and BullMQ handling job queuing.

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)

Log in to join the discussion and vote.

Log in

Related stories

0
ProgrammingDEV Community ·

Writer Realizes AI Has Become His First Instinct for Emotional Support Over Family

A writer reflecting on a late-night creative crisis found himself turning to ChatGPT before reaching out to his parents, prompting deeper self-examination about modern emotional habits. The AI responded with reassurance and a practical plan, mirroring the kind of comfort he had once received from his parents after a difficult exam. He noted that AI's constant availability, non-judgmental tone, and patience make it an instinctively easy first stop when life feels overwhelming. However, he drew a sharp distinction between AI, which retains information, and parents, who hold lived memories and genuine emotional history. The experience left him unsettled, raising broader questions about how quietly AI has reshaped where people instinctively seek comfort and connection.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Stylifyword Add-in Formats Word Documents Using Local AI, No Data Sent Externally

A new Microsoft Word add-in called Stylifyword aims to automate document formatting, going beyond text generation to apply heading hierarchies, list styles, and tables using a document's existing Word styles. The tool is paired with a companion desktop app that runs AI inference entirely on-device, meaning no data leaves the user's machine. All changes are output as tracked redlines in Word's Review tab, requiring user approval before any edits are applied. Stylifyword offers three deployment modes: fully local offline processing, a bring-your-own-key option compatible with Azure OpenAI, AWS Bedrock, or Anthropic, and a hosted pay-as-you-go tier. The add-in is positioned for use in regulated industries or environments with strict data confidentiality requirements.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Open WebUI testing reveals cleanup and runtime ownership gaps in Ota v1.6.24

Pressure-testing the Open WebUI repository against the Ota lifecycle tool exposed four significant weaknesses in how Ota handles runtime cleanup after successful proof runs. A key flaw was that native workloads launched via 'uv run open-webui serve' remained alive even after proof completed, causing CI pipeline blockages in GitHub Actions. A subsequent fix overcorrected by treating Docker Compose-managed runtimes the same as host-owned native processes, conflating two distinct cleanup ownership classes. Developers clarified that host-managed native workloads and adapter-owned Compose runtimes require separate teardown logic. These findings led to three concrete fixes in Ota v1.6.24, tightening proof teardown and aligning contract bootstrap truth with the repository's intended install configuration.

0 comments Read more at DEV Community
0
ProgrammingDEV Community ·

Developer Launches Golang Tutorial Series in Hinglish for Hindi-Speaking Learners

A developer going by the name CSM has started a tutorial series on the Go programming language, written in Hinglish — a mix of Hindi and English. The series aims to make Go accessible to Hindi-speaking audiences who may find English-only resources difficult to follow. The first installment covers the basics of Go, describing it as a compiled, garbage-collected language that runs faster than Python or JavaScript. The tutorial highlights Go's common use cases, including building APIs, web servers, cloud-native applications, CLI tools, and DevOps automation. The author plans to continue the series with more posts, each ending with a short Urdu or Hindi couplet as a signature touch.

0 comments Read more at DEV Community