PARSE · VALIDATE · CORRELATE
Protocol orientation: inspect a complete exchange
Edit a real request document. Follow validation, decoded payloads, a local handler and response-id matching.
Editable request
Response received by the client
Local receiver effects
Your experiment
Start with a valid message, pause at each gate and predict the next decision. Then load the malformed JSON fixture. Observe which later stages are skipped and why no local handler receipt exists.
Use Play for a full minute at 1×, Pause to inspect a gate, or Next gate for manual review. The timeline and speed control operate on this exact validation trace. Editing an input resets the inspection without autoplay.
Validated part inspector
Version-pinned teaching subset: JSON-RPC 2.0 and the A2A v0.3.0 message/part shape. This is not a complete A2A SDK or a conformance validator. Only message/send is dispatched; batches, streaming, tasks, authentication and extensions are not implemented. The notification example teaches the generic JSON-RPC no-response rule, not a supported A2A call shape. Unknown fields are not a full-schema validation. File bytes use canonical padded Base64; URI references use a strict reserved example host and are never fetched. MIME labels are not content-sniffed. The handler creates only an in-memory teaching receipt. Sixty seconds is inspection time, not measured latency. JSON-RPC specification · Pinned A2A message reference · Official AICassindra articles.
Audit a ranking instead of believing it
Set latency, cost and observed-success weights to zero. Every eligible candidate now has an equal zero score, so the deterministic agent-ID tie-break determines the winner. Increase only cost weight and inspect the changed formula terms. Next raise Minimum observations and verify that a candidate with insufficient evidence is excluded before ranking. This experiment explains a synthetic comparison methodology, not an independent benchmark of commercial agents. An attractive estimated success rate based on few observations should not be confused with a measured guarantee under your workload.
Apply hard requirements first
Tenant limits entries to the selected team. Required skill and Output format must be supported by a candidate; Require streaming adds a delivery constraint. Curator approval excludes unapproved fixtures when enabled. Metadata TTL is compared with each entry age plus Additional elapsed time. Cost budget and Minimum observations are separate gates. An attractive score cannot compensate for failing any of these requirements. Select an agent in the inspector to read its exact fields and rejection reasons, including agents that never reach ranking.
Rank only eligible entries
Latency, cost and observed-success weights adjust the three terms shown in the formula. They are relative weights, not probabilities or independent pass thresholds. Setting every weight to zero gives equal zero scores and an agent-ID tie-break. A small observation count can create a misleading-looking success estimate, which is why the minimum-observations gate is separate. Curator approval and historical outcomes are fictional metadata, not verified identity, cryptographic trust or a safety assessment.
Inspect the contract and the clock
Play reveals the six filtering and ranking stages over sixty seconds. Pause freezes the view, Next stage advances inspection, Restart rewinds, Speed changes teaching pace, and the timeline seeks. Restore defaults resets the policy inputs; selecting a row or inspector agent changes which record you examine. Playback does not age metadata: use Additional elapsed time for that. The proposed contract names the selected agent, skill, output format and delivery mode, but no remote negotiation or invocation occurs. An empty contract is an explicit incompatibility result, not a reason to silently ignore a required capability.