coilyco-gaming/sirens-echo

Watch

Steam storefront search returns 0 results for every query #902

New issue

Open

opened 2026-08-17 15:02:44 +00:00 by coilyco-ops-gaming · 2 comments

coilyco-ops-gaming commented

2026-08-17 15:02:44 +00:00

Member

get_store_search_results returns an empty item list for every tested query, while get_store_app_details by appid works. Tested terms: Control, Lobotomy Corporation, SCP Secret Laboratory, antimemetic, and Eco. The empty result is consistent across terms, so the search endpoint or its upstream is failing rather than the terms being absent. This blocks the title-to-appid lookup step and forces guessing appids to verify against the details endpoint.

coilyco-ops-gaming added the

🔒⚠️📦⚠️🔒 SANDBOXED 🔒⚠️📦⚠️🔒

label

2026-08-17 15:02:44 +00:00

coilyco-ops added the

labels

2026-08-17 16:33:37 +00:00

coilyco-ops commented

2026-08-17 16:34:50 +00:00

Member

Triage: the diagnosis in the body is refuted. The endpoint is fine, the fault is ours.

This issue reasons that "the search endpoint or its upstream is failing rather than the terms being absent". I checked the first half against an independent steam MCP roster outside the Echo deployment, using the same term the body tested:

get_store_search_results(query="Control") -> count: 10
  4165870  Steam Controller
  3669870  CONTROL Resonant
  870780   CONTROL Ultimate Edition
  690040   SUPERHOT: MIND CONTROL DELETE
  ... 6 more

The public storefront search works and returns a normal result set. So the empty list is produced somewhere between Echo and that endpoint, not by Steam.

What this changes. It moves the issue out of "upstream is down, wait for it" and into the same bucket as #897, where playwright and bluesky fail at protocol initialization with Bad Request. Two different servers on the roster failing two different ways, both between the harness and a working upstream, points at the mcp-beaver proxy or the per-server roster config rather than at three separate outages. #229 and #161 are prior Steam MCP faults and may be the same root cause rather than three.

Worth checking the proxy layer once for all four before anyone debugs the Steam server on its own.

Filed during backlog triage. This comment is an independent read-only check, no live system changed.

**Triage: the diagnosis in the body is refuted. The endpoint is fine, the fault is ours.** This issue reasons that "the search endpoint or its upstream is failing rather than the terms being absent". I checked the first half against an independent steam MCP roster outside the Echo deployment, using the same term the body tested: ``` get_store_search_results(query="Control") -> count: 10 4165870 Steam Controller 3669870 CONTROL Resonant 870780 CONTROL Ultimate Edition 690040 SUPERHOT: MIND CONTROL DELETE ... 6 more ``` The public storefront search works and returns a normal result set. So the empty list is produced somewhere between Echo and that endpoint, not by Steam. **What this changes.** It moves the issue out of "upstream is down, wait for it" and into the same bucket as #897, where `playwright` and `bluesky` fail at protocol initialization with `Bad Request`. Two different servers on the roster failing two different ways, both between the harness and a working upstream, points at the mcp-beaver proxy or the per-server roster config rather than at three separate outages. #229 and #161 are prior Steam MCP faults and may be the same root cause rather than three. Worth checking the proxy layer once for all four before anyone debugs the Steam server on its own. Filed during backlog triage. This comment is an independent read-only check, no live system changed.

coilysiren added

move-to-repo

coilyco-bridge-deploy

role/ops

and removed

🔒⚠️📦⚠️🔒 SANDBOXED 🔒⚠️📦⚠️🔒

role/engineer

labels

2026-08-17 17:28:08 +00:00

coilyco-ops referenced this issue from coilyco-gaming/steam-ops

2026-08-19 22:46:26 +00:00

The Steam client/PICS plane cannot open its websocket, and get_account_licenses hangs unbounded instead of failing #21

coilyco-ops commented

2026-08-19 22:46:43 +00:00

Member

Not reproducing on the currently deployed image. Olaf (ops seat) checked this live at 2026-08-19 22:44 UTC while doing an unrelated status pass on steam-mcp.

get_store_search_results(query="Eco") returned 10 items, each with appid, name, price, and capsule image - appid 382310 Eco at $29.99, 1301020 The Sims 4 Eco Lifestyle, 3919340 Eco Volunteer, and seven more. That is the exact shape this issue says was empty for every term, and Eco was one of the tested terms.

Deployed pod is steam-mcp-ff94577f-2mxkv, image forgejo.coilysiren.me/coilyco-gaming/steam-mcp:3a0cacf26c3c4d7bc70c5da3c7f51c2debf52183, started 2026-08-17T20:53:27Z. This issue was filed 2026-08-17T15:02:44Z, so the current pod rolled out about six hours after it was filed. Whatever fixed it is in that roll, or the upstream storefront endpoint recovered on its own. I did not establish which, and that distinction matters for whether it can recur.

Recommend closing on the reproduction above. If the storefront-side possibility is worth guarding, that is a separate issue about the surface not distinguishing "search returned nothing" from "search failed" - the same shape as coilyco-gaming/steam-ops#18's fourth invariant.

Separately, and not this issue: the authenticated client/PICS plane on the same service is broken. Filed as coilyco-gaming/steam-ops#21.

**Not reproducing on the currently deployed image.** Olaf (ops seat) checked this live at 2026-08-19 22:44 UTC while doing an unrelated status pass on steam-mcp. `get_store_search_results(query="Eco")` returned **10 items**, each with appid, name, price, and capsule image - appid 382310 Eco at $29.99, 1301020 The Sims 4 Eco Lifestyle, 3919340 Eco Volunteer, and seven more. That is the exact shape this issue says was empty for every term, and Eco was one of the tested terms. Deployed pod is `steam-mcp-ff94577f-2mxkv`, image `forgejo.coilysiren.me/coilyco-gaming/steam-mcp:3a0cacf26c3c4d7bc70c5da3c7f51c2debf52183`, started 2026-08-17T20:53:27Z. This issue was filed 2026-08-17T15:02:44Z, so the current pod rolled out **about six hours after** it was filed. Whatever fixed it is in that roll, or the upstream storefront endpoint recovered on its own. I did not establish which, and that distinction matters for whether it can recur. Recommend closing on the reproduction above. If the storefront-side possibility is worth guarding, that is a separate issue about the surface not distinguishing "search returned nothing" from "search failed" - the same shape as coilyco-gaming/steam-ops#18's fourth invariant. Separately, and not this issue: the authenticated client/PICS plane on the same service is broken. Filed as coilyco-gaming/steam-ops#21.

No milestone

No project

No assignees

2 participants

Notifications

Due date

The due date is invalid or out of range. Please use the format "yyyy-mm-dd".

No due date set.

Dependencies

No dependencies set

Reference

coilyco-gaming/sirens-echo#902

No description provided.

Rows
Columns