rhachet-brains-openai

rhachet brain.atom and brain.repl adapter for openai

install

npm install rhachet-brains-openai

note: this package bundles @openai/codex-sdk js for seamless cjs (e.g., jest) compatibility. vendor binaries come from the peer dep.

usage

import { genBrainAtom, genBrainRepl } from 'rhachet-brains-openai';
import { z } from 'zod';

// create a brain atom for direct model inference
const brainAtom = genBrainAtom({ slug: 'openai/gpt/4o-mini' });

// simple string output
const { output: explanation } = await brainAtom.ask({
  role: { briefs: [] },
  prompt: 'explain this code',
  schema: { output: z.string() },
});

// structured object output
const { output: { summary, issues } } = await brainAtom.ask({
  role: { briefs: [] },
  prompt: 'analyze this code',
  schema: { output: z.object({ summary: z.string(), issues: z.array(z.string()) }) },
});

// create a brain repl for agentic tasks
const brainRepl = genBrainRepl({ slug: 'openai/codex' });

// use ask() for read-only operations
const { output: { analysis } } = await brainRepl.ask({
  role: { briefs: [] },
  prompt: 'analyze this codebase',
  schema: { output: z.object({ analysis: z.string() }) },
});

// use act() for read+write operations
const { output: { proposal } } = await brainRepl.act({
  role: { briefs: [] },
  prompt: 'refactor this module',
  schema: { output: z.object({ proposal: z.string() }) },
});

continuation support

both atoms and repls support multi-turn conversations via episode continuation.

atoms

atoms support continuation via the openai responses api. pass the prior episode to continue the conversation:

// first call establishes context
const resultFirst = await brainAtom.ask({
  role: {},
  prompt: 'remember the secret word "PINEAPPLE42"',
  schema: { output: z.object({ content: z.string() }) },
});

// second call continues with prior context
const resultSecond = await brainAtom.ask({
  on: { episode: resultFirst.episode },
  role: {},
  prompt: 'what is the secret word I told you?',
  schema: { output: z.object({ content: z.string() }) },
});
// resultSecond.output.content contains "PINEAPPLE42"

repls

repls support continuation via the codex-sdk resumeThread() api. the episode exid contains the thread id prefixed with openai/codex/ for cross-supplier validation:

// first call starts a new thread
const resultFirst = await brainRepl.ask({
  role: {},
  prompt: 'remember the secret word "MANGO99"',
  schema: { output: z.object({ content: z.string() }) },
});
// resultFirst.episode.exid = "openai/codex/{threadId}"

// second call resumes the thread
const resultSecond = await brainRepl.ask({
  on: { episode: resultFirst.episode },
  role: {},
  prompt: 'what is the secret word I told you?',
  schema: { output: z.object({ content: z.string() }) },
});
// resultSecond.output.content contains "MANGO99"

limitations

• cross-supplier continuation is not supported: episodes from other brain suppliers (e.g., anthropic) cannot be used to continue openai conversations. this throws a BadRequestError.
• exid validation: the episode exid must start with openai/codex/ for repl continuation. this prevents accidental cross-supplier continuation attempts.

available brains

atoms (via genBrainAtom)

stateless inference without tool use.

| slug | model | context | cost (input) | cost (output) | cutoff | | ------------------------------ | ------------------- | ------- | ------------ | ------------- | ------- | | openai/gpt/5.2-instant | gpt-5.2-instant | 400K | $1.75/1M | $14/1M | 2025-08 | | openai/gpt/5.2-pro | gpt-5.2-pro | 400K | $21/1M | $168/1M | 2025-08 | | openai/gpt/5.2-thoughtful | gpt-5.2 | 400K | $1.75/1M | $14/1M | 2025-08 | | openai/gpt/codex/5.2 | gpt-5.2-codex | 400K | $10/1M | $40/1M | 2025-08 | | openai/gpt/5 | gpt-5 | 400K | $1.25/1M | $10/1M | 2024-09 | | openai/gpt/5-pro | gpt-5-pro | 272K | $15/1M | $120/1M | 2024-09 | | openai/gpt/5-thoughtful | gpt-5-thinking | 400K | $1.25/1M | $10/1M | 2024-09 | | openai/gpt/5-thoughtful-mini | gpt-5-thinking-mini | 400K | $0.25/1M | $2/1M | 2024-09 | | openai/gpt/5.1-instant | gpt-5.1-chat-latest | 400K | $1.25/1M | $10/1M | 2024-09 | | openai/gpt/5.1-thoughtful | gpt-5.1 | 400K | $1.25/1M | $10/1M | 2024-09 | | openai/gpt/codex/5.1-max | gpt-5.1-codex-max | 400K | $7.50/1M | $30/1M | 2024-09 | | openai/gpt/codex/5.1-mini | gpt-5.1-codex-mini | 400K | $3/1M | $12/1M | 2024-09 | | openai/gpt/4.1 | gpt-4.1 | 1M | $2/1M | $8/1M | 2024-06 | | openai/gpt/4.1-mini | gpt-4.1-mini | 128K | $0.40/1M | $1.60/1M | 2024-06 | | openai/gpt/4.1-nano | gpt-4.1-nano | 1M | $0.10/1M | $0.40/1M | 2024-06 | | openai/o/3 | o3 | 200K | $0.40/1M | $1.60/1M | 2024-06 | | openai/o/3-mini | o3-mini | 200K | $1.10/1M | $4.40/1M | 2024-06 | | openai/o/3-pro | o3-pro | 200K | $20/1M | $80/1M | 2024-06 | | openai/o/4-mini | o4-mini | 200K | $1.10/1M | $4.40/1M | 2024-06 | | openai/gpt/5-mini | gpt-5-mini | 400K | $0.25/1M | $2/1M | 2024-05 | | openai/gpt/4-turbo | gpt-4-turbo | 128K | $10/1M | $30/1M | 2023-12 | | openai/gpt/4o | gpt-4o | 128K | $2.50/1M | $10/1M | 2023-10 | | openai/gpt/4o-mini | gpt-4o-mini | 128K | $0.15/1M | $0.60/1M | 2023-10 | | openai/o/1 | o1 | 200K | $15/1M | $60/1M | 2023-10 | | openai/o/1-mini | o1-mini | 128K | $3/1M | $12/1M | 2023-10 | | openai/o/1-preview | o1-preview | 128K | $15/1M | $60/1M | 2023-10 |

repls (via genBrainRepl)

agentic code assistant with tool use via codex-sdk.

| slug | model | context | cost (input) | cost (output) | cutoff | | ------------------------ | ------------------ | ------- | ------------ | ------------- | ------- | | openai/codex | gpt-5.1-codex-max | 400K | $7.50/1M | $30/1M | 2024-09 | | openai/codex/mini | gpt-5.1-codex-mini | 400K | $3/1M | $12/1M | 2024-09 | | openai/codex/max | gpt-5.1-codex-max | 400K | $7.50/1M | $30/1M | 2024-09 | | openai/codex/5.1 | gpt-5.1-codex-max | 400K | $7.50/1M | $30/1M | 2024-09 | | openai/codex/5.2 | gpt-5.2-codex | 400K | $10/1M | $40/1M | 2025-08 | | openai/codex/mini/5.1 | gpt-5.1-codex-mini | 400K | $3/1M | $12/1M | 2024-09 | | openai/codex/max/5.1 | gpt-5.1-codex-max | 400K | $7.50/1M | $30/1M | 2024-09 |

sources

• openai api costs
• codex models documentation
• codex sdk documentation