Add Inwise OSS meeting-memory adapter - #347
Conversation
time-attack
left a comment
There was a problem hiding this comment.
I reproduced two blocking relay failures against head 785c8e3:
-
Unauthenticated pairing creation grows durable state without bound. After creating 300 pairings with a 1 ms TTL and allowing them to expire, all 300 records remained in the state file. One additional unauthenticated
POST /v1/pairingsincreased the count to 301 and rewrote the file from 97,429 to 97,750 bytes. Because expired records are never removed and every creation serializes the full collection, the public endpoint permits unbounded disk growth and increasingly expensive event-loop work. Please add durable expiry cleanup plus bounded admission/rate limiting. -
Accepted in-flight requests are lost on relay restart. I dispatched a CLI request and confirmed the edge received it with HTTP 200. After restarting the relay from the same persisted state, the edge response returned HTTP 404 (
Request is no longer pending) and the original CLI request lost its connection. The pairing credentials survive, but the broker queue and response state do not because they live only in process-local Maps. This violates the repository durability requirement for queued/in-flight work and prevents safe restarts or blue-green/multi-instance operation. Please move request lifecycle state to a durable shared broker/store.
The focused package build and all three integration tests pass, but they do not cover either failure mode.
|
Hi,
I made the changes needed. Please review and approve the PR, if everything
looks good. Really excited!
…On Wed, Aug 12, 2026 at 4:45 PM Sina Matian ***@***.***> wrote:
***@***.**** requested changes on this pull request.
I reproduced two blocking relay failures against head 785c8e3:
1.
*Unauthenticated pairing creation grows durable state without bound.*
After creating 300 pairings with a 1 ms TTL and allowing them to expire,
all 300 records remained in the state file. One additional unauthenticated POST
/v1/pairings increased the count to 301 and rewrote the file from
97,429 to 97,750 bytes. Because expired records are never removed and every
creation serializes the full collection, the public endpoint permits
unbounded disk growth and increasingly expensive event-loop work. Please
add durable expiry cleanup plus bounded admission/rate limiting.
2.
*Accepted in-flight requests are lost on relay restart.* I dispatched
a CLI request and confirmed the edge received it with HTTP 200. After
restarting the relay from the same persisted state, the edge response
returned HTTP 404 (Request is no longer pending) and the original CLI
request lost its connection. The pairing credentials survive, but the
broker queue and response state do not because they live only in
process-local Maps. This violates the repository durability requirement for
queued/in-flight work and prevents safe restarts or
blue-green/multi-instance operation. Please move request lifecycle state to
a durable shared broker/store.
The focused package build and all three integration tests pass, but they
do not cover either failure mode.
—
Reply to this email directly, view it on GitHub
<#347?email_source=notifications&email_token=BAOK6T7E23ULYQJCSNTXG2L5JT6Q5A5CNFSNUABKM5UWIORPF5TWS5BNNB2WEL2QOVWGYUTFOF2WK43UKJSXM2LFO4XTIOJSGIYDQNZWGUYKM4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2KYZTPN52GK4S7MNWGSY3L#pullrequestreview-4922087650>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/BAOK6TYRQURE3BM6W2LUNCD5JT6Q5AVCNFSNUABGKJSXA33TNF2G64TZHMYTGMJWGUZDOMZRHA5US43TOVSTWNJRGI2TMNZRHE2TPILWAI>
.
Triage notifications, keep track of coding agent tasks and review pull
requests on the go with GitHub Mobile for iOS
<https://github.com/notifications/mobile/ios/BAOK6T5INIOUDGLNICAU4TL5JT6Q5A5CNFSNUABKM5UWIORPF5TWS5BNNB2WEL2QOVWGYUTFOF2WK43UKJSXM2LFO4XTIOJSGIYDQNZWGUYKM4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2KUZTPN52GK4S7NFXXG>
and Android
<https://github.com/notifications/mobile/android/BAOK6T5H7NBR5ALND2ZTSV35JT6Q5A5CNFSNUABKM5UWIORPF5TWS5BNNB2WEL2QOVWGYUTFOF2WK43UKJSXM2LFO4XTIOJSGIYDQNZWGUYKM4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2K4ZTPN52GK4S7MFXGI4TPNFSA>.
Download it today!
You are receiving this because you authored the thread.Message ID:
***@***.***>
|
|
Hi!
Please let me know if there are queries. Would love to get this out the
door this week!
On Fri, Aug 14, 2026, 11:50 AM Shravani Vatti ***@***.***>
wrote:
… Hi,
I made the changes needed. Please review and approve the PR, if everything
looks good. Really excited!
On Wed, Aug 12, 2026 at 4:45 PM Sina Matian ***@***.***>
wrote:
> ***@***.**** requested changes on this pull request.
>
> I reproduced two blocking relay failures against head 785c8e3:
>
> 1.
>
> *Unauthenticated pairing creation grows durable state without bound.*
> After creating 300 pairings with a 1 ms TTL and allowing them to expire,
> all 300 records remained in the state file. One additional unauthenticated POST
> /v1/pairings increased the count to 301 and rewrote the file from
> 97,429 to 97,750 bytes. Because expired records are never removed and every
> creation serializes the full collection, the public endpoint permits
> unbounded disk growth and increasingly expensive event-loop work. Please
> add durable expiry cleanup plus bounded admission/rate limiting.
> 2.
>
> *Accepted in-flight requests are lost on relay restart.* I dispatched
> a CLI request and confirmed the edge received it with HTTP 200. After
> restarting the relay from the same persisted state, the edge response
> returned HTTP 404 (Request is no longer pending) and the original CLI
> request lost its connection. The pairing credentials survive, but the
> broker queue and response state do not because they live only in
> process-local Maps. This violates the repository durability requirement for
> queued/in-flight work and prevents safe restarts or
> blue-green/multi-instance operation. Please move request lifecycle state to
> a durable shared broker/store.
>
> The focused package build and all three integration tests pass, but they
> do not cover either failure mode.
>
> —
> Reply to this email directly, view it on GitHub
> <#347?email_source=notifications&email_token=BAOK6T7E23ULYQJCSNTXG2L5JT6Q5A5CNFSNUABKM5UWIORPF5TWS5BNNB2WEL2QOVWGYUTFOF2WK43UKJSXM2LFO4XTIOJSGIYDQNZWGUYKM4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2KYZTPN52GK4S7MNWGSY3L#pullrequestreview-4922087650>,
> or unsubscribe
> <https://github.com/notifications/unsubscribe-auth/BAOK6TYRQURE3BM6W2LUNCD5JT6Q5AVCNFSNUABGKJSXA33TNF2G64TZHMYTGMJWGUZDOMZRHA5US43TOVSTWNJRGI2TMNZRHE2TPILWAI>
> .
> Triage notifications, keep track of coding agent tasks and review pull
> requests on the go with GitHub Mobile for iOS
> <https://github.com/notifications/mobile/ios/BAOK6T5INIOUDGLNICAU4TL5JT6Q5A5CNFSNUABKM5UWIORPF5TWS5BNNB2WEL2QOVWGYUTFOF2WK43UKJSXM2LFO4XTIOJSGIYDQNZWGUYKM4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2KUZTPN52GK4S7NFXXG>
> and Android
> <https://github.com/notifications/mobile/android/BAOK6T5H7NBR5ALND2ZTSV35JT6Q5A5CNFSNUABKM5UWIORPF5TWS5BNNB2WEL2QOVWGYUTFOF2WK43UKJSXM2LFO4XTIOJSGIYDQNZWGUYKM4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2K4ZTPN52GK4S7MFXGI4TPNFSA>.
> Download it today!
> You are receiving this because you authored the thread.Message ID:
> ***@***.***>
>
|
|
Hi! |
|
Thanks for this, @Shravani889 — and sorry for the slow response. We're closing this PR. Per We don't want your idea to get lost, so we've captured it in #736 to triage on our side. Please feel free to open an |
|
Hi,
I was hopeful about seeing usage when this was first launched. We are
nearing the one month mark now. Please let's push until we can publish. I
added a note, please let me know what else is needed.
…On Fri, Aug 28, 2026 at 2:47 PM Sina Matian ***@***.***> wrote:
Closed #347 <#347>.
—
Reply to this email directly, view it on GitHub
<#347?email_source=notifications&email_token=BAOK6T676QQ3YWEUE4DLOH35MH4YTA5CNFSNUABQM5UWIORPF5TWS5BNNB2WEL2JONZXKZKFOZSW45CON52GSZTJMNQXI2LPNYXTGMBRHA4TMMZSG42TNJTSMVQXG33OU5WWK3TUNFXW5JLFOZSW45FMMZXW65DFOJPWG3DJMNVQ#event-30189632756>,
or unsubscribe
<https://github.com/notifications/unsubscribe-auth/BAOK6T2PDZB6NFUIBF5RXYT5MH4YTAVCNFSNUABGKJSXA33TNF2G64TZHMYTGMJWGUZDOMZRHA5US43TOVSTWNJRGI2TMNZRHE2TPILWAI>
.
Triage notifications, keep track of coding agent tasks and review pull
requests on the go with GitHub Mobile for iOS
<https://github.com/notifications/mobile/ios/BAOK6TYTZYA7AEKTP6NK4535MH4YTA5CNFSNUABQM5UWIORPF5TWS5BNNB2WEL2JONZXKZKFOZSW45CON52GSZTJMNQXI2LPNYXTGMBRHA4TMMZSG42TNJTSMVQXG33OU5WWK3TUNFXW5JLFOZSW45FKMZXW65DFOJPWS33T>
and Android
<https://github.com/notifications/mobile/android/BAOK6T3TZBA4EQ2MWBHEU3D5MH4YTA5CNFSNUABQM5UWIORPF5TWS5BNNB2WEL2JONZXKZKFOZSW45CON52GSZTJMNQXI2LPNYXTGMBRHA4TMMZSG42TNJTSMVQXG33OU5WWK3TUNFXW5JLFOZSW45FOMZXW65DFOJPWC3TEOJXWSZA>.
Download it today!
You are receiving this because you were mentioned.Message ID:
***@***.***>
|
Inwise OSS meeting layer for QM
We are providing Inwise OSS as a local meeting-memory layer for QM. It lets a personal QM agent search meetings, transcripts, people, upcoming meetings, and action items without requiring Inwise Cloud.
This PR includes the adapter under
integrations/inwise:inwiseCLI and QMtool.jsonSKILL.mdMeeting data remains in the user's local Inwise installation. Inwise provides conversational memory and action-ready context management.