Skip to main content
Glama

aw.speech.whisper_decode_step

Explicit bounded continuation of the REAL original Tiny/Base decoder for a whole accepted selected WAV. First call uses the SAME-session original encoder handle and null continuation. A nonterminal step returns only an opaque encrypted SAME-session decoder-pending handle, NOT a transcript; call this tool explicitly again with identical source/prepared/model/language/output, the CURRENT returned handle and a new operationId. Original complete token history, logits/stopping rules and full TXT/SRT remain unchanged; each call yields after at most64 original nonterminal steps for Tiny or32 for Base and keeps the existing30s work limit, session caps, TTL/admission and quotas. Rebuilds constant encoder KV with one original initial decoder forward and verifies retained exact hashes. Base reprojects six original values from the same pinned encoder source as needed; no new encoder run. Atomically replaces old private state; cancel/failure preserves the last accepted state. Consumed pending handles fail SUPERSEDED, not automatic heavy retry. No cache upload/download/inline tensors, model change, truncation, native CPU preemption, accuracy/all139/physical capacity/client disk access or provider acceptance claim. Auto retains original English default; original padded timestamps require review.

Tool call template

Replace each placeholder with a value matching the input schema. Only required parameters are included; add optional parameters as needed.

{
  "id": 1,
  "jsonrpc": "2.0",
  "method": "tools/call",
  "params": {
    "arguments": {
      "session": "<session>",
      "operationId": "<operationId>",
      "source": "<source>",
      "prepared": "<prepared>"
    },
    "name": "aw.speech.whisper_decode_step"
  }
}

Input Schema

NameRequiredDescriptionDefault
modelNoonnx-community/whisper-tiny
outputNotxt
sourceYesAn immutable input or result handle belonging to this session; never a URL/path/base64.
sessionYesPrivate short-lived session capability returned by aw.session.open. Never publish or log it.
languageNoauto
preparedYes
operationIdYes
continuationNo

Schema Changelog

Changes observed during successful MCP inspections.

  1. First observed

TDQS

B3.2/5.0
Behavior4/5

Does the description disclose side effects, auth requirements, rate limits, or destructive behavior?

Annotations only mark this as a non-read-only, non-destructive, closed-world operation; the description adds substantial behavior on top, including the 64/32-step yield ceiling, the retained 30s work limit, session caps, TTL/admission and quotas. It also discloses state semantics that annotations cannot express: old private state is atomically replaced, cancel/failure preserves the last accepted state, and consumed handles fail SUPERSEDED rather than being retried automatically.

Agents need to know what a tool does to the world before calling it. Descriptions should go beyond structured annotations to explain consequences.

Conciseness2/5

Is the description appropriately sized, front-loaded, and free of redundancy?

The operative instruction — call again with the current handle and a new operationId — is one clause inside a long run-on paragraph, so the definition is not front-loaded. Capitalized emphasis words (REAL, SAME, CURRENT, NOT) and a tail of negative disclaimers about caching, preemption and provider acceptance dilute the sentences that actually matter for invoking the tool.

Shorter descriptions cost fewer tokens and are easier for agents to parse. Every sentence should earn its place.

Completeness3/5

Given the tool's complexity, does the description cover enough for an agent to succeed on first attempt?

There is no output schema, so the description is the only source of return-value information; it states what a nonterminal step returns but only implies that a terminal step yields the transcript and TXT/SRT. For an eight-parameter, multi-call stepping tool with low schema coverage and no annotations beyond safety hints, the definition covers most of the hard cases but leaves the success-path result and non-SUPERSEDED failure modes underspecified.

Complex tools with many parameters or behaviors need more documentation. Simple tools need less. This dimension scales expectations accordingly.

Parameters2/5

Does the description clarify parameter syntax, constraints, interactions, or defaults beyond what the schema provides?

Schema description coverage is only 25% across eight parameters, so the description carries most of the burden, and it leaves continuation — the most consequential parameter — without any schema description, explaining it only indirectly through the call sequence. It does clarify that source/prepared/model/language/output must stay identical across steps and that auto retains the original English default, but it says nothing about the format or meaning of prepared, session, or operationId beyond what the regex patterns already imply.

Input schemas describe structure but not intent. Descriptions should explain non-obvious parameter relationships and valid value ranges.

Purpose4/5

Does the description clearly state what the tool does and how it differs from similar tools?

The description states a specific verb and resource: a bounded continuation of the Tiny/Base decoder over an already-accepted WAV, where nonterminal calls return an opaque handle rather than a transcript. That is far more than a restatement of the name and lets an agent distinguish this stepping tool from a one-shot transcription tool, though the purpose is buried under dense jargon and the sibling names are never referenced.

Agents choose between tools based on descriptions. A clear purpose with a specific verb and resource helps agents select the right tool.

Usage Guidelines3/5

Does the description explain when to use this tool, when not to, or what alternatives exist?

Sequencing guidance is unusually explicit: the first call uses the same-session encoder handle with null continuation, and every later call must repeat identical source/prepared/model/language/output plus the CURRENT handle and a NEW operationId. What is missing is routing guidance — nothing says when to choose this stepwise path versus a direct transcription or an export sibling, so the agent must infer that from context.

Agents often have multiple tools that could apply. Explicit usage guidance like "use X instead of Y when Z" prevents misuse.

Try in Browser

Glama MCP Gateway

Add one secure layer between your agents and this server.

Resources