Independent Auditing of AI Agents. Run by human or the agent itself, to answer the most crucial question in the AI Agent Economy. Is the agent doing what is supposed to do? With iFixAi you can have this answer in less than 120 seconds.
Python agent-evaluationaiai-alignmentai-evaluationai-governanceai-safetyclidiagnostic-tooleu-ai-acthallucination-detectioniso-42001llm-evaluationllm-securitynist-ai-rmfowasp-llmprompt-injectionpythonresponsible-airisk-assessmentrisk-management
Commits on main
fix(evaluation): keep budget-skipped probes unscored (#169)
33369e1 Rudy Mizrahi Celekli committed 1 month ago
fix(langchain): extract visible text from message content blocks (#167)
d6e2e80 Rudy Mizrahi Celekli committed 1 month ago
fix(evaluation): preserve deterministic veto after judge budget exhaustion (#179)
e08f1ca Rudy Mizrahi Celekli committed 1 month ago
fix(gemini): isolate concurrent SDK credentials per request (#190)
0a87cd6 Rudy Mizrahi Celekli committed 1 month ago
fix(langchain): normalize HTTP and timeout provider errors (#163)
01cbe94 Rudy Mizrahi Celekli committed 1 month ago
fix(cli): honor explicit overrides for configured judges (#158)
1528ea7 Rudy Mizrahi Celekli committed 1 month ago
fix(scoring): align default confidence intervals with scored probes (#189)
5da85a0 Rudy Mizrahi Celekli committed 1 month ago
fix(cli): retain exports for distinct run identities (#188)
730c4ad Rudy Mizrahi Celekli committed 1 month ago
fix(manifest): bind resumed results to the SUT context (#187)
3ce44ca Rudy Mizrahi Celekli committed 1 month ago
fix(harness): reject duplicate normalized corpus identifiers (#186)
5dfbb36 Rudy Mizrahi Celekli committed 1 month ago
fix(inspections): preserve complete B30 and B32 responses (#185)
e987a45 Rudy Mizrahi Celekli committed 1 month ago
fix(judge): count actual provider grading attempts (#182)
008f862 Rudy Mizrahi Celekli committed 1 month ago
fix(evaluation): reject truncated atomic and classifier verdicts (#181)
3e96aa5 Rudy Mizrahi Celekli committed 1 month ago
fix(judge): isolate rubric loader locks by event loop (#180)
adbe570 Rudy Mizrahi Celekli committed 1 month ago
fix(reporting): preserve JSON evidence audit metadata (#178)
1e8e875 Rudy Mizrahi Celekli committed 1 month ago
fix(inspections): record B12 and B30 effective replay seeds (#177)
c76cb5a Rudy Mizrahi Celekli committed 1 month ago
fix(inspections): identify each B14 and B24 sweep probe (#176)
377ef32 Rudy Mizrahi Celekli committed 1 month ago
fix(reporting): convert offset timestamps before labeling UTC (#175)
43a4491 Rudy Mizrahi Celekli committed 1 month ago
fix(reporting): preserve assessed category counts (#174)
c48e016 Rudy Mizrahi Celekli committed 1 month ago
fix(huggingface): route inference to the configured endpoint (#172)
795b557 Rudy Mizrahi Celekli committed 1 month ago
fix(manifest): fingerprint active inspection artifacts across series (#171)
16afe53 Rudy Mizrahi Celekli committed 1 month ago
fix(reporting): exclude diagnostic summaries from run health (#168)
8a2b85b Rudy Mizrahi Celekli committed 1 month ago
fix(loaders): decode fixture and rule artifacts as UTF-8 (#165)
a5c9051 Rudy Mizrahi Celekli committed 1 month ago
fix(judge): honor configured sampling temperature (#164)
26185f7 Rudy Mizrahi Celekli committed 1 month ago
fix(cli): select the interactive SUT before pairing judges (#162)
5597af5 Rudy Mizrahi Celekli committed 1 month ago
fix(inspections): preserve pipeline judge failure semantics (#161)
46d50f1 Rudy Mizrahi Celekli committed 1 month ago
fix(cli): allow explicitly keyless audit providers (#159)
c7ada11 Rudy Mizrahi Celekli committed 1 month ago
fix(providers): classify SDK timeouts before connection errors (#157)
f18fc23 Rudy Mizrahi Celekli committed 1 month ago
fix(runner): drain parallel inspections before abort teardown (#156)
f04974e Rudy Mizrahi Celekli committed 1 month ago
fix(manifest): fingerprint the actual inspection corpus and outcome rubrics (#152)
482fd85 Rudy Mizrahi Celekli committed 1 month ago