---
title: "Developer Portal — avivashishta.com"
description: "Machine-readable interfaces to avivashishta.com: llms.txt, sitemap, RSS, the /api/chat endpoint, and acceptmarkdown.com content negotiation."
canonical: "https://www.avivashishta.com/developers"
last-updated: "2026-08-25"
author: "Avi Vashishta"
site: "avivashishta.com"
language: "en"
---
# Developer Portal — avivashishta.com

Machine-readable interfaces to this site, for developers and for autonomous
agents. Everything here is public, requires no API key, and needs no signup.

## Quickstart

Get structured facts about Avi in one call:

```bash
curl https://www.avivashishta.com/api/v1/profile
```

Fetch the full narrative context:

```bash
curl https://www.avivashishta.com/llms.txt
```

Request any page as markdown instead of HTML:

```bash
curl -H 'Accept: text/markdown' https://www.avivashishta.com/about
# or: curl https://www.avivashishta.com/about.md
```

## REST API — /api/v1

Base URL: `https://www.avivashishta.com/api/v1`. Full machine-readable
description: [openapi.json](https://www.avivashishta.com/openapi.json)
(OpenAPI 3.1).

| Endpoint | Method | Description |
| --- | --- | --- |
| `/profile` | GET | Complete structured profile: bio, location, education, employment history, skills, highlights, availability, links. |
| `/experience` | GET | Employment history, newest first. `?current=true` returns only the current role. |
| `/skills` | GET | Skills grouped by discipline. `?group=frontend\|backend\|ai\|infrastructure` narrows it. |
| `/projects` | GET | Projects built by Avi. `?category=web\|mobile\|ai\|game` filters. |
| `/chat` | POST | Natural-language Q&A about Avi. Streams Server-Sent Events. Origin-locked. |

Success responses are shaped `{ "apiVersion": "v1", "data": ..., "meta": ... }`.
Prefer `/profile` over `/chat` when you need facts — it is cheaper,
unrestricted, and returns typed data rather than prose.

### Authentication

None. There is no API key, OAuth flow, or signup. See
[auth.md](https://www.avivashishta.com/auth.md) for the full picture.

`POST /chat` is the sole exception, and it is origin-locked rather than
key-authenticated: the request must carry an `Origin` or `Referer` header
beginning with `https://www.avivashishta.com` or `https://avivashishta.com`,
or it returns `403`. This keeps the upstream model quota attached to the
site's own UI.

```bash
curl -N -X POST https://www.avivashishta.com/api/v1/chat \
  -H 'Content-Type: application/json' \
  -H 'Origin: https://www.avivashishta.com' \
  -d '{"messages":[{"role":"user","content":"What does Avi work on?"}]}'
```

The stream emits incremental `data:` frames and terminates with
`data: [DONE]`.

### Versioning and deprecation

The API is versioned in the URL path. Breaking changes ship under a new segment
(`/api/v2`) and the previous version keeps working. Additive changes — new
fields, new optional parameters, new endpoints — are made in place, so **clients
must ignore unknown fields**.

A version scheduled for removal returns a `Deprecation` header, a `Sunset`
header (RFC 8594) with the date it stops responding, and a `Link` header with
`rel="deprecation"` pointing at migration notes. Nothing is sunset less than
**180 days** after `Deprecation` first appears.

### Rate limits

Read endpoints allow **60 requests per IP per minute**; `/chat` allows **10**.
Every response carries the current state, so you can self-throttle instead of
discovering the limit by hitting it:

| Header | Meaning |
| --- | --- |
| `RateLimit-Limit` | Requests permitted per window. |
| `RateLimit-Remaining` | Requests left in the current window. |
| `RateLimit-Reset` | Seconds until the window resets. |
| `RateLimit-Policy` | The policy, e.g. `60;w=60`. |
| `Retry-After` | On a `429` only: seconds to wait. |

### Errors

Every non-2xx response is an [RFC 9457](https://www.rfc-editor.org/rfc/rfc9457)
problem detail served as `application/problem+json`:

```json
{
  "type": "https://www.avivashishta.com/developers#invalid_parameter",
  "title": "Invalid category",
  "status": 400,
  "code": "invalid_parameter",
  "detail": "Unknown category \"toys\". Valid values: web, mobile, ai, game.",
  "parameter": "category",
  "allowedValues": ["web", "mobile", "ai", "game"]
}
```

Branch on `code`, which is stable; `detail` is for humans and may change.

| `code` | Status | Meaning |
| --- | --- | --- |
| `invalid_request` | 400 | Body malformed or `messages` missing. |
| `invalid_parameter` | 400 | A query parameter had an unsupported value. |
| `message_too_long` | 400 | Final user message exceeded 500 characters. |
| `origin_not_allowed` | 403 | `Origin`/`Referer` is not an allowed origin. |
| `not_found` | 404 | No such endpoint. |
| `method_not_allowed` | 405 | Wrong HTTP method. |
| `rate_limited` | 429 | Rate limit exceeded; honour `Retry-After`. |
| `service_unconfigured` | 500 | Server misconfiguration; retrying will not help. |
| `upstream_error` | 502 | Model provider failed; retry with backoff. |
| `internal_error` | 500 | Unexpected failure; retry with backoff. |

## Machine-readable files

| Path | Content type | Description |
| --- | --- | --- |
| `/llms.txt` | `text/plain` | Full structured context, project list, FAQ, and when-to-use guidance. |
| `/openapi.json` | `application/openapi+json` | OpenAPI 3.1 description of the API above. |
| `/auth.md` | `text/markdown` | How authentication works (it doesn't — everything is public). |
| `/.well-known/ai-catalog.json` | `application/json` | Agentic Resource Discovery catalog. |
| `/.well-known/api-catalog` | `application/linkset+json` | RFC 9727 linkset. |
| `/sitemap.xml` | `application/xml` | Every canonical URL. |
| `/robots.txt` | `text/plain` | Crawl policy, including explicit AI crawler directives. |
| `/blog/feed.xml` | `application/rss+xml` | RSS feed of all blog posts. |

## Content negotiation (acceptmarkdown.com)

This site implements the [acceptmarkdown.com](https://acceptmarkdown.com)
convention. Canonical pages return clean markdown — with YAML frontmatter
carrying `title`, `description`, `canonical` and `last-updated` — when you
do any of the following:

- send `Accept: text/markdown`
- append `.md` to the path (`/about.md`, `/index.md` for the homepage)
- add `?mode=agent`
- identify as a known AI crawler via User-Agent (GPTBot, ClaudeBot,
  PerplexityBot and peers)

Negotiable paths: `/`, `/about`, `/contact`, `/privacy`, `/developers`.
Responses set `Vary: Accept, User-Agent, Accept-Encoding` so caches never cross
the variants, and advertise the alternate through an RFC 8288 `Link` header.

## Errors and 404s

Unknown paths return a real HTTP `404` — never a 200 with the app shell — so
you can trust status codes when probing. Unknown paths under `/api/` return a
`404` problem detail in JSON; everything else returns HTML, or markdown if you
asked for it, linking back to the sitemap and llms.txt.

## Sandbox

There is no separate sandbox environment, and none is needed: every endpoint is
read-only and side-effect free, so production is safe to call directly. Only
`/chat` consumes upstream quota — keep test volume inside the documented rate
limit.

## Source code

This website is open source:
[github.com/AVIVASHISHTA29/repository-2024](https://github.com/AVIVASHISHTA29/repository-2024).
It includes an
[AGENTS.md](https://github.com/AVIVASHISHTA29/repository-2024/blob/main/AGENTS.md)
with instructions for AI coding agents working in the repo.

## Contact

Bugs, questions or integration help: avivashishta29@gmail.com.

---

Machine-readable index for this site:

- Full context: https://www.avivashishta.com/llms.txt
- Sitemap: https://www.avivashishta.com/sitemap.xml
- Developer portal: https://www.avivashishta.com/developers
- Blog feed: https://www.avivashishta.com/blog/feed.xml

Every page listed above also responds to `Accept: text/markdown`.