API7.ai (@api7_ai) header
API7.ai (@api7_ai) profile picture

https://teapawt.org/t.co/eLZTfeIA5Y is an #opensource software infrastructure company that empowers #APIManagement and analysis.

Global
Joined February 2023
Using Claude on Bedrock or Vertex AI? AISIX can add prompt-cache markers for supported direct Claude models. Enable the setting, then verify cache creation and reads with your provider. Setup: docs.api7.ai/ai-gateway/traf… #AISIX #PromptCaching
6
Gateway CPU spiking on a remote host? Use API7 Gateway Enterprise Diagnostic Agents to capture a worker CPU profile from the Dashboard. Review capture quality before drawing conclusions. Setup: docs.api7.ai/api7-gateway/ob… #APIGateway #Observability
6
Ready to try GPT‑6.2 without changing your gateway URL? Configure Astra and Sol as model aliases on one AISIX Responses endpoint. Set up OpenAI with AISIX: docs.api7.ai/ai-gateway/gett… #AISIX #AIGateway
17
No key → 401. Valid + in quota → upstream (200 here). Over quota → configured 429. See APISIX pair key-auth with limit-count on one route (per IP, per instance). Key auth: apisix.apache.org/docs/apisi… Rate limiting: apisix.apache.org/docs/apisi…
8
Does seeing a token count in your request log mean you were charged again? In AISIX Enterprise 1.9.0, an exact cache hit can still log tokens. It skips the upstream call and records zero billed cost. Configure response caching docs.api7.ai/ai-gateway/traf… #AISIX #AIGateway
6
How do you roll out v2 gradually? This APISIX demo uses traffic-split to start small, inspect responses, widen the share, and restore v1. Counts are one local sample, not a guaranteed ratio. Docs: apisix.apache.org/docs/apisi…
6
Upstream 200 with a non-empty top-level error. In our local Completions test with a simulated upstream, AISIX 1.8.0 returned 200; 1.9.0 returned 502. Some covered errors may return 4xx instead. Details: docs.api7.ai/ai-gateway/rele… #AISIX
17
Claude Fable 5.2 on AISIX. Map claude-fable-5-2 to your model alias through the Anthropic integration. Connect Claude with AISIX: docs.api7.ai/ai-gateway/prov… #AISIX #Claude
2
One request. One working Apache APISIX route. See GET /hello move through a local gateway: match the path, choose an upstream, create the route, then verify the response. Official guide: apisix.apache.org/docs/apisi… #ApacheAPISIX
5
Same GraphQL depth, different quota cost. APISIX 3.19 charged 4 vs 13 units for first:1 and first:10 in our local test. An over-quota request stopped before the simulated upstream. Setup → docs.api7.ai/hub/graphql-lim… #APISIX
1
A client sends GET /orders. Who decides what happens next? Follow one request through Apache APISIX. Watch the gateway choose a backend, apply rate limits, and expose metrics. Then follow a request to an LLM. Try your first route. apisix.apache.org/docs/apisi… #ApacheAPISIX
11
Same app alias. Gateway-managed defaults. We changed AISIX’s default temperature from 1.0 to 0.7 and replayed the identical request. The upstream value changed; the app alias stayed stable. Real gateway · simulated upstream. Guide → docs.api7.ai/ai-gateway/mode… #AISIX
1
Which team drove your AI bill? AISIX 1.7 tags show gateway-recorded cost, tokens & requests by department. Demo data shown. docs.api7.ai/ai-gateway/clou… #AISIX
1
Gzipped JSON should still meet your API contract. New in API7 Enterprise 3.10.9: request-decompress runs before OpenAPI validation. Our local demo shows valid data passing and invalid data rejected. Plugin guide → docs.api7.ai/hub/request-dec… #APIGateway
1