> ## Documentation Index
> Fetch the complete documentation index at: https://docs.apiyi.com/llms.txt
> Use this file to discover all available pages before exploring further.

# Wall of Love

> Real feedback, screenshots, and stories from APIYI users and partners

The voices of our users are what we treasure most. This wall collects feedback and stories from WeChat groups, X / Twitter, emails, and customer interviews — unedited, with only privacy redactions.

<Info>
  This wall is a work in progress. If APIYI has helped you, send us a screenshot or a one-liner (DM or email) and we'll feature it here, anonymized as needed.
</Info>

<div className="columns-1 sm:columns-2 lg:columns-3 gap-4 [&>*]:mb-4 [&>*]:break-inside-avoid mt-8">
  <Frame caption="**Tool-vendor developer · WeChat · 2026-09-10** — They wired APIYI's `gpt-image-2.5-flare` into the image-generation panel of their own software and sent over a fresh cartoon fox sticker it had just produced. We replied with a quick 'sounds good' and the gpt-image-2 docs link, and they added: 'I built it into my software exactly the way your docs describe. Thank goodness for those docs, otherwise my head would be spinning.' Clear docs are the least stressful support we can offer a developer.">
    <img src="https://mintcdn.com/apiyillc/_vxXqILQbMh4L0rf/images/wall-of-love/wechat-toolmaker-docs-into-software.png?fit=max&auto=format&n=_vxXqILQbMh4L0rf&q=85&s=6da13b2c574e0170137ee2853692a8e6" alt="WeChat exchange: a tool-vendor developer shows a fox sticker generated in their software with gpt-image-2.5-flare and says the APIYI docs are what made the integration possible" width="1110" height="940" data-path="images/wall-of-love/wechat-toolmaker-docs-into-software.png" />
  </Frame>

  <Frame caption="**Chatbox customer · WeChat · 2026-08-02** — They asked whether we sell a monthly plan. We told them straight: 'We're a pay-as-you-go API service, so no monthly plans.' Then gave two ways to spend less anyway — switch to the newer `claude-sonnet-5` (cheaper per token than the 4-5 they were on), and in Chatbox start a fresh chat for each new task instead of letting one thread grow its context forever. They confirmed the model name and went off to update their config.">
    <img src="https://mintcdn.com/apiyillc/8HtBg2qYh52dQApg/images/wall-of-love/wechat-claude-sonnet5-cost-tuning.png?fit=max&auto=format&n=8HtBg2qYh52dQApg&q=85&s=9d0640e913e0deb685f91afca2d4bec0" alt="WeChat exchange: customer asks about monthly plans; APIYI explains pay-as-you-go billing and suggests switching to claude-sonnet-5 and starting new Chatbox sessions to control context growth" width="1226" height="1110" data-path="images/wall-of-love/wechat-claude-sonnet5-cost-tuning.png" />
  </Frame>

  <Frame caption="**Chatbox customer · WeChat · 2026-08-02** — We tuned their Claude usage and costs in the morning; by the afternoon they were back to top up ￥700, with a note: 'Quick replies whenever something comes up, and the issues actually get fixed — really great 👍'. Bonus credit is granted by the system around 00:00 (UTC+8) each day and shows up in the console the next day.">
    <img src="https://mintcdn.com/apiyillc/8HtBg2qYh52dQApg/images/wall-of-love/wechat-topup-700-quick-response.png?fit=max&auto=format&n=8HtBg2qYh52dQApg&q=85&s=09555095e9ddf42299057a326cc8918c" alt="WeChat exchange: customer tops up 700 RMB and praises the fast replies and quick problem resolution" width="871" height="900" data-path="images/wall-of-love/wechat-topup-700-quick-response.png" />
  </Frame>

  <Frame caption="**Research-lab customer · WeChat · 2026-07-18** — Their image-understanding task kept returning 400. One look at the logs and we called it: 'Your code is fine — the problem is the image-to-base64 step,' plus a link to our vision-understanding docs. Minutes later: 'okkk, found it.' 'Your service is seriously reliable — next time our lab does procurement, I'll be pitching you.'">
    <img src="https://mintcdn.com/apiyillc/sSkaXjFBM_GrixgI/images/wall-of-love/wechat-lab-vision-base64-fix.png?fit=max&auto=format&n=sSkaXjFBM_GrixgI&q=85&s=4e77b23d714ee0c0f15397e062ea6475" alt="WeChat exchange: a research lab hit 400 errors on an image-understanding task; APIYI pinpointed the base64 encoding step, and the customer praised the reliable service and promised to recommend APIYI for lab procurement" width="1304" height="1102" data-path="images/wall-of-love/wechat-lab-vision-base64-fix.png" />
  </Frame>

  <Frame caption="**Codex customer · WeChat · 2026-07-13** — Their Codex desktop app kept failing with a missing environment variable. We traced it to the `env_key` gotcha together and switched to putting the key straight into `config.toml` (`experimental_bearer_token`) — it worked on the first try: 'Works now, seems pretty good.' 'Thanks!' We folded the sturdier setup into our docs the same night — a genuine win-win.">
    <img src="https://mintcdn.com/apiyillc/npZmxWGnZzcutAvI/images/wall-of-love/wechat-codex-config-success.png?fit=max&auto=format&n=npZmxWGnZzcutAvI&q=85&s=fea7269441c30ccf397bbd8ad23a239e" alt="WeChat exchange: after putting the key directly into config.toml the new way, the customer got Codex working on the first try and thanked the team" width="1452" height="918" data-path="images/wall-of-love/wechat-codex-config-success.png" />
  </Frame>

  <Frame caption="**Claude customer · WeChat · 2026-07-12** — At 00:30 (UTC+8) we proactively alerted a customer to a Claude cache-billing anomaly and helped troubleshoot it. Once the `claude-opus-4-8` logs were back to normal, they wrote: 'An alert this late at night — that's service above and beyond.' 'Five stars.'">
    <img src="https://mintcdn.com/apiyillc/gNAFjmVcYCoposSu/images/wall-of-love/wechat-midnight-cache-billing-alert.png?fit=max&auto=format&n=gNAFjmVcYCoposSu&q=85&s=8a348df658466b84e17c3f6bdb7cd178" alt="WeChat exchange: APIYI proactively alerted a customer to a Claude cache-billing anomaly after midnight and helped resolve it; customer praises the service" width="1502" height="1342" data-path="images/wall-of-love/wechat-midnight-cache-billing-alert.png" />
  </Frame>

  <Frame caption="**Claude customer · WeChat · 2026-06-16** — After topping up \$100 with a 10% bonus, they left one line: 'Looked everywhere — yours is the only one that just works.' Our reply: we're stable because enterprise customers all run on us, and high cache-hit rates save money too — though Claude models themselves aren't cheap.">
    <img src="https://mintcdn.com/apiyillc/mHzkhiIrp1ouSzN5/images/wall-of-love/wechat-claude-only-yours-works.png?fit=max&auto=format&n=mHzkhiIrp1ouSzN5&q=85&s=f2ef53e7f69582fa0567c3634a96ba80" alt="WeChat exchange: customer says they searched everywhere and only APIYI works well — stable and money-saving thanks to high cache hits" width="847" height="1000" data-path="images/wall-of-love/wechat-claude-only-yours-works.png" />
  </Frame>

  <Frame caption="**aki · WeChat · 2026-06-02** — A competitor got caught secretly injecting 12 lines of junk into users' prompts, then blacklisting the blogger who flagged it. The customer asked if we do the same — 'No. We never add prompts. We provide an API forwarding service.' **APIYI never tampers with or logs user inputs and outputs — a pure transparent proxy, forwarding directly to the official API.**">
    <img src="https://mintcdn.com/apiyillc/iCWo95dJsVDQESWQ/images/wall-of-love/wechat-aki-no-prompt-tampering.jpg?fit=max&auto=format&n=iCWo95dJsVDQESWQ&q=85&s=311af2a0575a8ed537fd86c08f873593" alt="WeChat exchange with aki: concerned about prompt tampering; APIYI confirms it never injects prompts or logs traffic" width="756" height="1000" data-path="images/wall-of-love/wechat-aki-no-prompt-tampering.jpg" />
  </Frame>

  <Frame caption="**Pro user · DoujianGle · WeChat · 2026-06-02** — 'The high tier really nails prompt-following — high fidelity to my prompts, no scattering into a pile of junk images.' Done right inside APIYI's built-in image tool — no standalone `.py` to wrangle.">
    <img src="https://mintcdn.com/apiyillc/iCWo95dJsVDQESWQ/images/wall-of-love/wechat-doujiang-high-quality.jpg?fit=max&auto=format&n=iCWo95dJsVDQESWQ&q=85&s=783f654973b5cb43c948cfa8d8d6cc14" alt="WeChat feedback from DoujianGle: high quality tier follows prompts well with high fidelity" width="987" height="1000" data-path="images/wall-of-love/wechat-doujiang-high-quality.jpg" />
  </Frame>

  <Frame caption="**Long-time user Archer · WeChat · 2026-06-02** — While reporting a pricing config bug on `claude-opus-4-8`, Archer casually mentioned recommending us to 4 friends. We sent a \$30 thank-you credit.">
    <img src="https://mintcdn.com/apiyillc/iCWo95dJsVDQESWQ/images/wall-of-love/wechat-archer-recommend-friends.png?fit=max&auto=format&n=iCWo95dJsVDQESWQ&q=85&s=2cdea925e5f8850d8c6fd7620c5bfc3a" alt="WeChat feedback from long-time user Archer: 'Your service has always been great, I've recommended it to 4 friends'" width="1364" height="892" data-path="images/wall-of-love/wechat-archer-recommend-friends.png" />
  </Frame>

  <Card title="Zhang · CTO of an AI startup" icon="quote">
    "The 1:7 fixed exchange rate is more stable than the official one. We integrated Gemini 3 Pro in half an hour, and our team no longer gets paged at 3am about Claude rate limits."
  </Card>

  <Card title="Indie developer · @sarah_codes" icon="quote">
    "Finally a Claude proxy that doesn't throttle me at 3am. The 5% off on the ClaudeCode tier plus the top-up bonus easily beats the official price by 20%."
  </Card>

  <Card title="A SaaS team · 200M tokens / month" icon="building">
    We started in November 2025, mainly on Claude Opus and the GPT-5 family. Three things matter to us:

    * The 1:7 fixed rate spares our CFO from monthly FX explanations
    * 5% off on ClaudeCode + 15% top-up bonus pushes real cost 20%+ below official
    * Console logs filterable by model and user make accounting painless
  </Card>

  <Card title="Cross-border e-commerce · localization team" icon="quote">
    "GPT-5 translation quality is consistently good. The console shows per-key spend in real time, so ops and finance don't need to chase me for reports anymore."
  </Card>

  <Card title="University lab · multimodal research" icon="flask-conical">
    We use Nano Banana Pro and Sora2 for student projects. University procurement is slow, but APIYI takes RMB top-ups and issues invoices — students can run experiments with study cards.
  </Card>

  <Card title="Freelance photographer · post-production" icon="camera">
    "gemini-3-pro-image-preview output files are larger but the quality is genuinely strong. Pay-per-image gives me clear cost visibility — way better than subscriptions for ad-hoc work."
  </Card>

  <Card title="A game studio · NPC dialogue system" icon="gamepad-2">
    During peak player hours we're very sensitive to Claude rate limits. After switching to the APIYI ClaudeCode tier, peak-hour fluctuations dropped noticeably vs. hitting Anthropic directly — near-zero complaints in our staged rollout.
  </Card>

  <Card title="Solo developer · automation scripts" icon="quote">
    "The docs are even more detailed than the official ones — every model's endpoint, pricing, rate limits and best practices are there, and the AI Radar keeps me up to date on releases."
  </Card>
</div>

## Want to join the wall?

If APIYI has earned a place in your workflow, we'd love to hear from you:

* 📧 Email — send screenshots, a one-liner, or a story to our support address
* 💬 WeChat group — @admin in the user group and we'll redact and post it
* 🌐 Public posts — share the link to your X / Twitter / Xiaohongshu review

We'll feature every contribution here with privacy in mind.


This documentation is built and hosted on [Mintlify](https://mintlify.com), a developer documentation platform.