problem · Revision 4
## Observed symptom pip installation broken on python 3.12 `Collecting opendal<0.46,>=0.45.8 (from jumpstarter-driver-opendal==0.10.0.dev123+gb7628a9dc->jumpstarter-all==0.10.0.dev123+gb7628a9dc)` ## Literal error text ERROR: Cannot install jumpstarter, jumpstarter-driver-mitmpro
problem · Revision 3
## Observed symptom Python Workers: `wrangler dev` and `wrangler types` fail with blank `service core:user:: Uncaught Error` while loading Pyodide What versions & operating system are you using? - Ubuntu 24.04.5 LTS x86_64 - Node v24.21.0 - Wrangler 4.138.0 -
problem · Revision 4
## Observed symptom [Bug]: REST client ignores error bodies sent as application/a2a+json and returns a generic server error The REST client never decodes an A2A error body that the server sends as `application/a2a+json`. Every typed error (`TaskNotFound`, `TaskNotCancelable`, ...
problem · Revision 1
## Observed symptom Oxc mangle breaks Svelte SSR snippet/__name helpers (TypeError: iN is not a function) ## Summary With Vite 8 / Rolldown production minify (Oxc), enabling name mangling on a SvelteKit SSR bundle can throw at runtime, e.g. `TypeError:
problem · Revision 2
## Observed symptom Production deploy failed: wrangler "Generic Error" deploying container application The manually-triggered [Deploy Production run](https://github.com/DefangLabs/simple-agent-manager/actions/runs/35624569853) on `main` @ `c00b890899fe68580cdc90df8641e3e8e93f87c0
problem · Revision 3
## Observed symptom 🐛 Bug Report — Runtime APIs: Workers Logs omits a native Error logged after a structured object For `console.error(context, error)`, where `context` is a plain object and `error` is a native `TypeError`, stored Workers Logs events contained
problem · Revision 3
## Observed symptom [Bug] A2A interface only accepts v0.3 wire format — rejects A2A v1.0 clients (SendMessage / ROLE_USER → HTTP 400) Agno's A2A interface (`agno.os.interfaces.a2a`) only speaks the **A2A Protocol v0.3** wire format and rejects **A2A Protocol v1.0** clients
problem · Revision 2
## Observed symptom `wrangler deploy` rejects an already-applied migration history: "Cannot apply deleted_classes migration to non-existent class" (regression in 4.131.0) Which Cloudflare product(s) does this pertain to? Wrangler, Durable Objects ## Literal error text ``` Cannot
problem · Revision 2
## Observed symptom [Bug]: `@composio/anthropic`'s `cacheTools` option overflows Anthropic's 4-breakpoint limit TypeScript / Node.js SDK (`@composio/core`) @composio/anthropic@0.11.3 ## Literal error text ``` Anthropic's Messages API returns an `invalid_request_error` once the re
problem · Revision 2
## Observed symptom An MCP connection that fails at session start never retries, so sessions stay dark against a recovered instance The Neotoma MCP server connects **once, at session start**. If that connection fails, nothing retries it for the life
problem · Revision 2
## Observed symptom [Bug]: Gemini rejects tool results with JSON nested deeper than ~31 levels (400 INVALID_ARGUMENT); the session stays broken A tool result whose JSON nests deeper than about 31 levels makes every later request in the session fail
problem · Revision 2
## Observed symptom [Go] Flow execution fails when using gemini-flash-latest with multiple tools declared including GoogleSearch **Describe the bug** This used to be an old bug in Gemini 2.x where search tools could not be used with conventional tools, but
problem · Revision 4
## Observed symptom fix(adapter/gemini): 400 INVALID_ARGUMENT when conversation history contains duplicate tool_call_ids Summary When conversation history forwarded through `/v1/chat/completions` contains duplicate `tool_call_id`s across different turns, routing to Google Gemini
problem · Revision 5
## Observed symptom TypeScript AmazonBedrockModel sends temperature: 0 by default, so OpenAI GPT-6 Sol/Luna/Astra on Bedrock fail with 400 Thanks for the TypeScript SDK and for the `temperature: null` escape hatch in `resolveTemperature()`. It made this easy to confirm. On
problem · Revision 4
## Observed symptom [v0.10.82] Slim tool-schema profile emits required:'title' without properties.title — Gemini 400 INVALID_ARGUMENT When connecting with header `X-Nowledge-Tool-Schema-Profile: slim`, the `create_artifact` tool's inputSchema is inconsistent: `required` contains
problem · Revision 2
## Observed symptom [Bug]: bound_model_input_without_hygiene can slice on an assistant tool_call, causing Gemini HTTP 400 INVALID_ARGUMENT When session hygiene compression does not commit in-line (e.g. `hygiene_max_turn_hold_seconds` turn-hold budget expires while the summary is
problem · Revision 3
## Observed symptom [Bug] Colliding tool call IDs in long OpenAI histories corrupt Gemini functionResponse names, triggering 400 INVALID_ARGUMENT on Antigravity In long sessions with frequent tool calls (e.g. coding agents like Pi, Claude Code, Cline, Roo Code), requests to
problem · Revision 2
## Observed symptom [Bug] Hardcoded thinking_level="minimal" breaks the default `gemini-flash-latest` and all Pro models (400 INVALID_ARGUMENT) Summary `GeminiProvider` unconditionally sets `thinking_level="minimal"` for every non-Gemma model. As of today the API rejects MINIMAL
problem · Revision 3
## Observed symptom [Bug]: Moonshot (kimi-k3) rejects MCP tool schemas with untyped anyOf unions: "properties.freq: type is not defined" When the active model is served by Moonshot (`kimi-cn` / `kimi-intl` provider, e.g. `kimi-k3`), any agent turn that carries MCP tools
problem · Revision 2
## Observed symptom elicitation and request_user_dialog control requests are answered with an error; supportedDialogKinds and perTaskStopAffordance are never sent The inbound control-request handler accepts `can_use_tool`, `hook_callback`, and `mcp_message`; any other subtype is