Python · Java · Agent / RAG
Build local Agent prototypes for Text2SQL, knowledge retrieval, and tool calling, and check execution through parameter validation, access tests, and task evaluation.
[0:00] This is the first line of the embedded subtitle test. [0:02] This is subtitle track line two
Actual test output · no upload endpoint
Use existing text layers or subtitles first, keep source references, and retry only unfinished items.
Mechanism evaluation · fixed reference SQL
A local Agent prototype for natural-language data queries, with verified read-only safeguards. A separate evaluation tests schema recall, join paths, and failure cases.
SVG delivery workflowExtract topic-specific sentences from annual reports and export sentence-level results in a Windows workflow the client could run.
Saved result · translated demo material
Separate proposals from confirmed decisions, trace tasks to their sources, and leave unstated deadlines open.
Trace tool calls, file handling, and saved deliverables to locate failures in compound tasks.
Review evaluation methods and failure patterns ↗The local tool-calling module preserves sources and asks for confirmation before saving. Review tested cancellation, expiry, and repeat-confirmation paths.
Review confirmation replay ↗Use tool traces to diagnose invalid arguments, blocked tool calls, and timeouts in a Java integration.
API & setup ↗Change the work goal to see a different reading priority for the same technical book.
Offline app ↗Compare ideas across books around one question, check passages, and remove unsupported links before export.
Review cross-book evidence ↗Distinguish missing evidence from a policy violation, and explain what to supply before rule review.
Two scenarios ↗