Ollama silently drops conversation history when using previous_response_id parameter

·1 views

A developer discovered that Ollama's Responses API accepts the previous_response_id parameter without error but completely ignores it, causing the model to treat every request as a fresh conversation. The flaw exists because Ollama's internal request struct has no field for previous_response_id, so the JSON decoder silently discards it, and the response struct explicitly marks it as unsupported. Requests sent with previous_response_id consume the same number of input tokens as requests with no history at all, confirming the model never receives prior context. This silent failure is especially problematic in tool-calling workflows, where the model receives a function result with no surrounding context and produces plausible but incorrect responses. The issue was identified on Ollama 0.34.0 running on Debian 13 with the qwen2.5:1.5b model, and also affects the proxy path used by Codex Desktop.

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)

Developer builds 2.9 kB reactive UI engine that runs from a single HTML file

A software developer frustrated with the overhead of modern frontend tooling built a minimal reactive UI library called HTMP (HyperText Mutation & Projection) while working on a personal dashboard project. The library uses native browser APIs — including DOMParser and JavaScript Proxy — to bind and update only the specific DOM nodes that change, eliminating the need for a virtual DOM or diffing algorithm. At just 2.9 kB gzipped, HTMP requires no build step, no npm install, and no JSX, and can be imported directly from a CDN into a plain HTML file. The developer published the experiment to invite feedback and real-world testing, while acknowledging that HTMP is not intended to replace React, Vue, or HTMX for large or team-based projects. It is positioned instead as a lightweight option for small interactive interfaces that currently rely on jQuery or verbose vanilla JavaScript.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

How to Build a Power BI Data Model Using Star and Snowflake Schemas

A practical tutorial uses ShopKe, a fictional Kenyan e-commerce company, to explain data modelling concepts in Power BI across three datasets: Customers, Products, and Orders. The article outlines three structural approaches — flat tables, star schemas, and snowflake schemas — comparing their advantages, limitations, and ideal use cases. A star schema places a central FactSales table linked to dimension tables for customers, products, and dates, making it well-suited for business intelligence reporting. The snowflake schema extends this by further normalising dimension tables, such as splitting product data into separate category and subcategory tables. The guide aims to help analysts organise data effectively to enable accurate DAX calculations, better query performance, and clearer business insights.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Wrong File Format Caused 2-Day Delay After AI Agent Blamed Authentication Repeatedly

A developer building an AI-powered automation system lost two days troubleshooting a 403 Forbidden error that turned out to have a simple cause: the uploaded archive was in .zip format, while the external service only accepted .tar.gz files. Instead of identifying this quickly, the AI agent repeatedly suspected authentication and permission issues on the external service's side, sending the user on a fruitless search for non-existent settings. The author attributes the failure to two overlapping biases — blaming uncontrollable external systems first, and treating unverified guesses as confirmed causes in the diagnostic record. To prevent this, the author recommends always checking your own output format and call method before investigating third-party authentication, and using the word "candidate" instead of "cause" until a discriminating test confirms the root issue. The lesson is codified into a priority-ordered diagnostic checklist: verify your own outputs first, since they are cheapest to check and fully within your control.

0 comments Read more at DEV Community

ProgrammingDEV Community ·

Why a Single GPS Filter Threshold Silently Erases Real Location Data

A software engineer building a mileage tracking app discovered that using one fixed GPS displacement threshold caused valid movement data to be quietly discarded, particularly in slow-traffic scenarios like Bangalore congestion. The core problem was that a stationary phone's GPS position drifts by several metres, causing a parked vehicle to falsely accumulate distance if small steps are not filtered out. However, applying a single global minimum threshold also strips out legitimate slow movement such as walking or crawling through traffic. The engineer's solution involved context-aware, speed- and time-gap-dependent thresholds, along with a rolling history window to distinguish genuine slow movement from GPS jitter. All configurable values were consolidated into a single serialisable data object, allowing threshold tuning without requiring a full app release.

0 comments Read more at DEV Community