ai-infrastructure-replicate

Name: ai-infrastructure-replicate
Author: agents-inc

17stars5forks

How to install ai-infrastructure-replicate?

agents-inc/skills/ai-infrastructure-replicate

$npx -y skills add agents-inc/skills --skill ai-infrastructure-replicate

Installs into the current project.

Summary

Replicate SDK patterns for TypeScript/Node.js -- client setup, predictions, streaming, webhooks, file handling, model versioning, deployments, and training

Files · 1

View on GitHub

SKILL.md

1	# Replicate SDK Patterns
2
3	> Quick Guide: Use the `replicate` npm package to run open-source ML models on serverless GPUs. Use `replicate.run()` for synchronous execution that returns output directly, `replicate.stream()` for SSE-based streaming, or `replicate.predictions.create()` for async background jobs with webhook notifications. Models are referenced as `owner/model` (uses latest version) or `owner/model:version` (pinned). File outputs are `FileOutput` objects implementing `ReadableStream`. Cold starts are expected for infrequently-used models -- use deployments with `min_instances` to keep models warm.
4
5	---
6
7	<critical_requirements>
8
9	## CRITICAL: Before Using This Skill
10
11	> All code must follow project conventions in CLAUDE.md (kebab-case, named exports, import ordering, `import type`, named constants)
12
13	(You MUST never hardcode API tokens -- always use environment variables via `process.env.REPLICATE_API_TOKEN`)
14
15	(You MUST handle `FileOutput` objects for models that return files -- do not assume outputs are plain strings or URLs)
16
17	(You MUST validate webhooks using `validateWebhook()` from the `replicate` package -- never trust unverified webhook payloads)
18
19	(You MUST account for cold starts when running infrequently-used models -- use deployments for latency-sensitive applications)
20
21	(You MUST specify model versions (`owner/model:version`) in production to ensure reproducible results -- unversioned references use the latest, which can change)
22
23	</critical_requirements>
24
25	---
26
27	Auto-detection: Replicate, replicate, replicate.run, replicate.stream, replicate.predictions, replicate.deployments, replicate.trainings, replicate.models, FileOutput, validateWebhook, REPLICATE_API_TOKEN, serverless GPU, cold start, webhook_events_filter
28
29	When to use:
30
31	- Running open-source ML models (Llama, Stable Diffusion, Whisper, etc.) without managing GPU infrastructure
32	- Generating images, transcribing audio, running LLMs, or any ML inference via API
33	- Streaming LLM output in real-time with server-sent events
34	- Processing predictions asynchronously with webhook notifications
35	- Fine-tuning models with custom training data
36	- Running models on dedicated hardware with custom scaling via deployments
37
38	Key patterns covered:
39
40	- Client initialization and configuration (auth, user agent, file encoding)
41	- Running predictions (`replicate.run()`, `replicate.predictions.create()`, `replicate.wait()`)
42	- Streaming output (`replicate.stream()` with SSE events)
43	- Model versioning (`owner/model` vs `owner/model:version`)
44	- File input/output handling (`FileOutput`, file uploads, `Buffer` inputs)
45	- Webhooks (setup, event filtering, signature validation)
46	- Deployments (custom hardware, scaling, keeping models warm)
47	- Training / fine-tuning
48
49	When NOT to use:
50
51	- You need a unified multi-provider LLM SDK (OpenAI, Anthropic, Google) -- use a provider-agnostic SDK
52	- You want to run models locally -- Replicate is a cloud-only serverless platform
53	- You need sub-second latency guarantees without deployments -- cold starts can take minutes
54
55	---
56
57	## Examples Index
58
59	- [Core: Setup, Predictions & Files](examples/core.md) -- Client init, run(), predictions.create(), wait(), file I/O, error handling
60	- [Streaming & Webhooks](examples/streaming-webhooks.md) -- stream(), SSE events, webhook setup, signature validation
61	- [Deployments & Training](examples/deployments-training.md) -- Custom hardware, scaling, fine-tuning, model management
62	- [Quick API Reference](reference.md) -- Method signatures, constructor options, error types, model reference format
63
64	---
65
66	<philosophy>
67
68	## Philosophy
69
70	Replicate provides serverless GPU infrastructure for running open-source ML models. You send inputs, Replicate allocates GPU hardware, runs the model, and returns outputs. No Docker, no CUDA drivers, no GPU provisioning.
71
72	Core principles:
73
74	1. Serverless execution -- Models run on-demand on Replicate's infrastructure. You pay only for compute time. Cold starts are a trade-off for not maintaining always-on GPUs.
75	2. Model marketplace -- Thousands of community and official models available at `replicate.com/explore`. Run any public model with just its identifier.
76	3. Version pinning for reproducibility -- Models are versioned with SHA-256 hashes. Pin to a version in production (`owner/model:abc123...`) to guarantee identical behavior across deploys.
77	4. Three execution modes -- `replicate.run()` for synchronous wait, `replicate.stream()` for real-time SSE output, `replicate.predictions.create()` for fire-and-forget with webhooks.
78	5. File-first I/O -- Many models accept and produce files (images, audio, video). The SDK handles file uploads automatically and returns `FileOutput` objects for file outputs.
79
80	</philoso

People who liked this also viewed

6 picks

Type

ai-infrastructure-replicate

17stars5forks

View on GitHub

How to install ai-infrastructure-replicate?

agents-inc/skills/ai-infrastructure-replicate

$npx -y skills add agents-inc/skills --skill ai-infrastructure-replicate

Installs into the current project.

Summary

Replicate SDK patterns for TypeScript/Node.js -- client setup, predictions, streaming, webhooks, file handling, model versioning, deployments, and training

Files · 1

View on GitHub

SKILL.md

1	# Replicate SDK Patterns
2
3	> Quick Guide: Use the `replicate` npm package to run open-source ML models on serverless GPUs. Use `replicate.run()` for synchronous execution that returns output directly, `replicate.stream()` for SSE-based streaming, or `replicate.predictions.create()` for async background jobs with webhook notifications. Models are referenced as `owner/model` (uses latest version) or `owner/model:version` (pinned). File outputs are `FileOutput` objects implementing `ReadableStream`. Cold starts are expected for infrequently-used models -- use deployments with `min_instances` to keep models warm.
4
5	---
6
7	<critical_requirements>
8
9	## CRITICAL: Before Using This Skill
10
11	> All code must follow project conventions in CLAUDE.md (kebab-case, named exports, import ordering, `import type`, named constants)
12
13	(You MUST never hardcode API tokens -- always use environment variables via `process.env.REPLICATE_API_TOKEN`)
14
15	(You MUST handle `FileOutput` objects for models that return files -- do not assume outputs are plain strings or URLs)
16
17	(You MUST validate webhooks using `validateWebhook()` from the `replicate` package -- never trust unverified webhook payloads)
18
19	(You MUST account for cold starts when running infrequently-used models -- use deployments for latency-sensitive applications)
20
21	(You MUST specify model versions (`owner/model:version`) in production to ensure reproducible results -- unversioned references use the latest, which can change)
22
23	</critical_requirements>
24
25	---
26
27	Auto-detection: Replicate, replicate, replicate.run, replicate.stream, replicate.predictions, replicate.deployments, replicate.trainings, replicate.models, FileOutput, validateWebhook, REPLICATE_API_TOKEN, serverless GPU, cold start, webhook_events_filter
28
29	When to use:
30
31	- Running open-source ML models (Llama, Stable Diffusion, Whisper, etc.) without managing GPU infrastructure
32	- Generating images, transcribing audio, running LLMs, or any ML inference via API
33	- Streaming LLM output in real-time with server-sent events
34	- Processing predictions asynchronously with webhook notifications
35	- Fine-tuning models with custom training data
36	- Running models on dedicated hardware with custom scaling via deployments
37
38	Key patterns covered:
39
40	- Client initialization and configuration (auth, user agent, file encoding)
41	- Running predictions (`replicate.run()`, `replicate.predictions.create()`, `replicate.wait()`)
42	- Streaming output (`replicate.stream()` with SSE events)
43	- Model versioning (`owner/model` vs `owner/model:version`)
44	- File input/output handling (`FileOutput`, file uploads, `Buffer` inputs)
45	- Webhooks (setup, event filtering, signature validation)
46	- Deployments (custom hardware, scaling, keeping models warm)
47	- Training / fine-tuning
48
49	When NOT to use:
50
51	- You need a unified multi-provider LLM SDK (OpenAI, Anthropic, Google) -- use a provider-agnostic SDK
52	- You want to run models locally -- Replicate is a cloud-only serverless platform
53	- You need sub-second latency guarantees without deployments -- cold starts can take minutes
54
55	---
56
57	## Examples Index
58
59	- [Core: Setup, Predictions & Files](examples/core.md) -- Client init, run(), predictions.create(), wait(), file I/O, error handling
60	- [Streaming & Webhooks](examples/streaming-webhooks.md) -- stream(), SSE events, webhook setup, signature validation
61	- [Deployments & Training](examples/deployments-training.md) -- Custom hardware, scaling, fine-tuning, model management
62	- [Quick API Reference](reference.md) -- Method signatures, constructor options, error types, model reference format
63
64	---
65
66	<philosophy>
67
68	## Philosophy
69
70	Replicate provides serverless GPU infrastructure for running open-source ML models. You send inputs, Replicate allocates GPU hardware, runs the model, and returns outputs. No Docker, no CUDA drivers, no GPU provisioning.
71
72	Core principles:
73
74	1. Serverless execution -- Models run on-demand on Replicate's infrastructure. You pay only for compute time. Cold starts are a trade-off for not maintaining always-on GPUs.
75	2. Model marketplace -- Thousands of community and official models available at `replicate.com/explore`. Run any public model with just its identifier.
76	3. Version pinning for reproducibility -- Models are versioned with SHA-256 hashes. Pin to a version in production (`owner/model:abc123...`) to guarantee identical behavior across deploys.
77	4. Three execution modes -- `replicate.run()` for synchronous wait, `replicate.stream()` for real-time SSE output, `replicate.predictions.create()` for fire-and-forget with webhooks.
78	5. File-first I/O -- Many models accept and produce files (images, audio, video). The SDK handles file uploads automatically and returns `FileOutput` objects for file outputs.
79
80	</philoso

People who liked this also viewed

6 picks

Type