The latest articles on DEV Community by Mohamed Osama (@mohamed_osama_9b102cb789b). https://dev.to/mohamed_osama_9b102cb789b https://media2.dev.to/dynamic/image/width=90,height=90,fit=cover,gravity=auto,format=auto/https:%2F%2Fdev-to-uploads.s3.us-east-2.amazonaws.com%2Fuploads%2Fuser%2Fprofile_image%2F1912122%2F578365c4-c6b0-4569-8503-6f3437e2d431.jpg https://dev.to/mohamed_osama_9b102cb789b en Mohamed Osama Sat, 07 Nov 2026 18:31:05 +0000 https://dev.to/mohamed_osama_9b102cb789b/when-a-tool-goes-silent-after-execution-retry-or-verify-first-27m0 https://dev.to/mohamed_osama_9b102cb789b/when-a-tool-goes-silent-after-execution-retry-or-verify-first-27m0 <p>The discussion under <a href="https://dev.to/mohamed_osama_9b102cb789b/gemini-tool-calling-why-completion-claims-need-verifiable-results-3kl9">my previous post</a> got me thinking about a more specific situation than simply saying "done": what if execution may have succeeded, but we cannot verify it?</p> <p>Imagine an assistant sends a request to create a record, then the connection drops before the response arrives. The missing response alone does not tell the user whether the record was created. Retrying immediately could create a duplicate.</p> <p>This is a hypothetical example for discussion, not a failure I have reproduced in Gemini.</p> <p>For me, a useful message in that situation would be:</p> <blockquote> <p>I sent the request, but I couldn't confirm the result. I'll check whether the record exists before retrying.</p> </blockquote> <p>I'd like to hear from people building agents that use tools: how do you connect an execution attempt to a check of its outcome? How do you decide which operations are safe to repeat and which need verification or user input?</p> <p>An error or success message alone isn't enough here. The user needs to know the next step without risking a duplicate operation.</p> discuss Mohamed Osama Wed, 04 Nov 2026 19:07:37 +0000 https://dev.to/mohamed_osama_9b102cb789b/gemini-tool-calling-why-completion-claims-need-verifiable-results-3kl9 https://dev.to/mohamed_osama_9b102cb789b/gemini-tool-calling-why-completion-claims-need-verifiable-results-3kl9 <p>In my personal experience using Gemini in tool-enabled workflows, I have encountered responses that present a task as completed when I cannot verify the intended outcome. I want to discuss that gap constructively with other developers.</p> <p>A confident completion message affects what a user does next: trust the result, retry an operation, or investigate a failure. If the message is not grounded in the actual tool result, the user ends up doing the verification work anyway.</p> <h2> Scope of this feedback </h2> <p>This is qualitative feedback from my own experience, not a benchmark or a claim about every Gemini model or integration. I am not attaching a minimal reproducible example, exact model/version details, or request/response traces here. I therefore cannot isolate the cause to the model rather than the application, orchestration layer, or tool integration.</p> <h2> The distinction I want applications to make </h2> <ul> <li> <strong>Proposed:</strong> the assistant describes an action.</li> <li> <strong>Attempted:</strong> a tool call was issued.</li> <li> <strong>Tool-reported success:</strong> the tool returned a success response.</li> <li> <strong>Verified outcome:</strong> the resulting state was checked and matches the requested outcome.</li> </ul> <p>Those states should not all be described as "done." If a tool fails, access is unavailable, or the final state cannot be checked, the response should say so clearly.</p> <p>For workflows that change something, I would like applications to check the resulting state where practical before reporting completion. When only part of a task succeeds, the response should identify that part and what remains unresolved.</p> <h2> Questions for developers </h2> <p>How do you keep completion claims grounded in tool results in Gemini function-calling workflows? What patterns have worked for ambiguous tool responses, partial success, and verification after an action?</p> <p>I would also welcome suggestions for the most useful details to include in a reproducible follow-up report.</p> <p>I shared the same concern on the <a href="https://discuss.ai.google.dev/t/gemini-tool-calling-feedback-on-unverified-task-completion-claims/198215" rel="noopener noreferrer">Google AI Developers Forum</a>.</p> <p>I am interested in practical ways to make these workflows more reliable and their status messages more accurate.</p> agents discuss gemini llm Mohamed Osama Sun, 31 May 2026 21:04:28 +0000 https://dev.to/mohamed_osama_9b102cb789b/building-hermes-financial-agent-an-explainable-ai-copilot-for-egx-investors-bg0 https://dev.to/mohamed_osama_9b102cb789b/building-hermes-financial-agent-an-explainable-ai-copilot-for-egx-investors-bg0 <h2> Overview </h2> <p>I built <strong>Hermes Financial Agent</strong> — an AI-powered financial assistant for investors in the Egyptian Exchange (EGX).</p> <p>The goal: not just calculate portfolio value, but explain risks in a transparent, auditable way.</p> <h2> Features </h2> <ul> <li>Real EGX market data via Yahoo Finance</li> <li>Portfolio tracking and valuation</li> <li>Daily financial reports</li> <li>Telegram-user portfolio isolation</li> <li>Cached quote fallback for resilience</li> <li>Explainable risk insights</li> </ul> <h2> Explainable Risk Intelligence </h2> <p>Unlike traditional trackers, Hermes surfaces:</p> <ul> <li>Portfolio concentration risk</li> <li>Stale market data exposure</li> <li>Quote coverage percentage</li> <li>Valuation gaps from unavailable data</li> </ul> <p>Users understand <em>confidence level</em> behind their valuation — not just the number.</p> <h2> Technical Stack </h2> <ul> <li> <strong>Hermes Agent</strong> framework</li> <li>Python</li> <li>GitHub Actions (automated smoke testing)</li> <li>Offline-safe quote fallback architecture</li> </ul> <h2> Repository </h2> <p>🔗 <a href="https://github.com/Abozied1313/hermes-agent" rel="noopener noreferrer">GitHub Repository</a></p> <h2> Future Roadmap </h2> <ul> <li>News sentiment analysis</li> <li>Investment thesis tracking</li> <li>Market briefing generation</li> <li>Advanced financial reasoning agents</li> </ul> <h1> hermeschallenge </h1> python ai webdev programming