riivault_/ENTITY 4418
● LiveproductLLM frameworksrepo:mastra-ai/mastranpm:@mastra/core
Entity drill-down · last 90 days ·30d90d180d

Mastra

Hype · Discussion moved +2578.6% while downloads moved -21.8%. (as of 2026·08·12)

Narrow: 258 distinct voices behind 1500 mentions.

Who is talking
0.17
258 distinct voices across 1500 mentions

Distinct voices per mention. 1.00 means every mention came from a different person; a low number means a handful of people are driving the volume. Authors are counted once per day, so someone who posts across several days counts more than once over the whole window. The figure can overstate breadth, never understate it.

Where it is being said
github100%
hackernews0%

A wave of praise on a forum and a rapidly filling issue tracker both show up as “mentions up” until you see where they came from.

Agreement
Aligned
spread 0.55 · sides 39% even

Most opinion points the same way, so the average sentiment reads as a verdict rather than two sides cancelling out.

SRC: 1,562 mentions in window · net sentiment 0.23 · 3 adoption series · derived aggregates only

Mentions / dayall sources combined
06-0108-0108-12
latest 282 mentions06-01 .. 08-12
Sentiment curve0.23 current
06-0108-0108-12
latest 0.2306-01 .. 08-12
GitHub releases / dayadoption
05-1507-0108-12
latest 105-15 .. 08-12
GitHub stars (total)adoption
08-0608-0908-12
latest 27.1k08-06 .. 08-12
npm downloads / dayadoption
07-0707-2308-09
latest 105.8k07-07 .. 08-09
What people are asking forlatest 20 entries
What was saidKindTimesFirst / LastMomentum
editor.prompt.update() on pinned blocks causes cache poisoning that makes updates invisible to other processes while appearing successful to the updater.Bug108-12 → 08-12+0%
durableAgent.recover() does not re-announce recovered runs on the thread topic, leaving reconnecting clients without knowledge of recovery.Bug108-12 → 08-12+0%
Mastra ClickHouse client issues DDL statements without consuming the response stream, causing uncaught warnings and potential connection errors.Bug108-12 → 08-12+0%
User asks whether Mastra will officially support Valkey or rely on existing Redis compatibility.Feature request108-12 → 08-12+0%
Truthiness check treats an empty headers object as valid gateway auth, short-circuiting the getApiKey fallback.Bug108-12 → 08-12+0%
A regression in version 1.56.0 skips output processors when a provider stream fails.Bug108-12 → 08-12+0%
Dynamic workspace skill paths are stored in a mutable shared cache causing cross-request state leakage.Bug108-12 → 08-12+0%
WorkflowsPG.listWorkflowRuns() lacks default indexes causing repeated full scans on large tables.Pain point108-12 → 08-12+0%
Langfuse score export diverges from configured trace environment causing scores to silently disappear.Bug108-12 → 08-12+0%
Run.cancel() leaves span tree open when a step ignores abortSignal in Mastra.Bug108-12 → 08-12+0%
Expose per-message Signal metadata to Channel handlers for transport context like attachment references.Feature request108-12 → 08-12+0%
Add support for traceable redaction tokens with stable identifiers instead of a single flat redaction value to preserve referential structure in traces.Feature request108-12 → 08-12+0%
Anthropic 1-hour prompt-cache writes are exported at the 5-minute rate due to a missing TTL dimension in the cache write tracking.Bug108-12 → 08-12+0%
Langfuse scores are ingested into the wrong environment because the LangfuseExporter does not pass the environment parameter to the underlying client.Bug108-12 → 08-12+0%
Add a Gandr TTS voice provider to Mastra as a new @mastra/voice-gandr package following the @mastra/voice-openai shape.Feature request108-12 → 08-12+0%
The listWorkflowRuns method performs full table scans on mastra_workflow_snapshot due to missing database indexes, causing performance degradation on large datasets.Bug108-12 → 08-12+0%
Approval suspensions can be self-approved when the auto-resume system defaults to approved:true and accepts model-authored resumeData as the human's answer.Bug108-12 → 08-12+0%
Workspace skills resolver is cached on the singleton instance rather than per-request, causing concurrent requests to clobber each other's skill sets mid-turn.Bug108-12 → 08-12+0%
outputProcessors are skipped when the provider throws, preventing memory processors from running and the turn from being written.Bug108-11 → 08-11+0%
requestContextSchema types model a closed record but the runtime is an open map, forcing casts for runtime-only keys and breaking generic Agent assignability.Bug108-11 → 08-11+0%