Remove misc prompts

cognesy · Nov 11, 2024 · 718e5b0 · 718e5b0
1 parent a9b0265
commit 718e5b0
Show file tree

Hide file tree

Showing 6 changed files with 159 additions and 340 deletions.
diff --git a/.gitignore b/.gitignore
@@ -47,4 +47,5 @@ Thumbs.db
 /experimental/*
 /tmp/
 /temp/
-/prompts/misc/
+/prompts/misc/*
+/prompts/archive/*
diff --git a/docs/quickstart.mdx b/docs/quickstart.mdx
@@ -29,8 +29,7 @@ In your project directory, create a new PHP file `test-instructor.php`:
 
 ```php
 <?php
-$loader = require 'vendor/autoload.php';
-$loader->add('Cognesy\\Instructor\\', __DIR__ . '../../src/');
+require __DIR__ . '/vendor/autoload.php';
 
 use Cognesy\Instructor\Instructor;
 

diff --git a/notes/NOTES.md b/notes/NOTES.md
@@ -1,154 +1,155 @@
-# NOTES
-
-## High priority
-
-- Evals / eval framework
-  * execution level correctness metric
-- Add 'Output' section to each example, generate it and include in docs, so reader can see what they can expect
-- Logging via PSR-3
-- Schema abstraction layer - decouple names and descriptions from the model
-- Prompt optimization via TextGrad
-- Agents
-- Indexing to vector DB
-- CLI app
-- Fast/simple REST API server - compatible with OpenAI? 
-- Revise examples debugging - not sure if it works as expected (what does it demonstrate?)
-- Multiple tools with tool selection
-- Parallel tool calls
-- Export configuration to user folder / use external configuration
-- Full control over generated prompt (access to Script object processing)
-
-## Low priority
-
-- Batch API support (Gemini, OpenAI, Anthropic)
-- Gemini context caching
-- Generate unstructured, then format to structured - to improve reasoning
-
-
-# BACKLOG
-
-## Addon: Modules
-
-- Modules: Add module observability via events - currently no access to this info
-
-
-## Refactorings
-
-- More modular design - serialization, validation, transformation should be a configurable pipeline
-- Rework Events so they have toArray() method, make __toString() use it
-- Role - should be enum, not string?
-- String >> Array >> Class - for example: prompts (they should be classes)
-- Prompt - should be a class, not a string; prompt translates to Section/Messages; alt name: Instruction(s)
-
-## Infrastructure
-
-- Async mode
-- Fix response caching
-- Finish logging support
-- Document how to inject custom HTTP client
-- Better error messages
-- Documentation for logging
-- PSR-14 events - finish, demo how to plug custom dispatcher
-- PSR-11 container - finish, demo how to plug custom container
-- PSR-3 logger - finish, demo how to plug custom logger
-
-## Other
-
-- Sequences - validate only individual items, reject ONLY the invalid; allows progressive extraction in multiple stages
-- Use LLM to generate Example based on the class - just render object to JSON + add schema as an explanation - this should give the model enough info to come up with something that makes sense
-- Extract APIClient to a separate package?
-- Foreign client support - OpenAI PHP, Amazon Bedrock SDK, etc. - via some kind of adapters?
-- example of integration with Laravel/Livewire
-- Read rate limits from API responses
-- Rate limited API calls - wait for the limit to reset
-
-
-
-# Partially done
-
-- Better API for image / audio inputs
-- Make system, prompt, script etc. available for configuration by user
-- How to track API rate limits across multiple requests / parallel executions
-- Make using DocBlocks optional - it may not always to be desired to pass this info to LLM
-- Add super detailed tests of Module core functionality - esp. around input/output mappings
-- Validators / Deserializers / Transformers - chain of objects, not a single object
-- API Client: Clean up predefined models, prices, etc.
-
-
-
-
-# Brain dump
-
-- Streaming JSON parser - https://github.com/kelunik/streaming-json
-- Retry - https://github.com/kelunik/retry
-- Rate limiter - https://github.com/kelunik/rate-limit
-- Check for broken links - https://github.com/kelunik/link-check
-- Async flow execution - https://github.com/darkwood-com/flow
-- Models to implement JsonSerializable: https://www.php.net/manual/en/jsonserializable.jsonserialize.php / https://www.sitepoint.com/use-jsonserializable-interface/
-- Support JsonException for serialization / deserialization errors - https://www.php.net/manual/en/class.jsonexception.php
-- ValueError - https://www.php.net/manual/en/class.valueerror.php
-- Task runner - https://robo.li/
-- Hub >> Laravel Zero - https://laravel-zero.com/
-- DSPy next steps: eval, optimize, compile
-- Add more modules: XoT, RAG, ReAct, etc.
-- RAG - how to handle multiple VDB providers?
-- Test validation in modules - provide an example
-- Parallel execution of modules (e.g. a la Laravel jobs?)
-- Moderation endpoint support
-- Git/GitHub integration module to allow easy automation
-- Data mapping module(s) for easier data transformations
-- How to handle dynamic module graph definition + visualization
-- JSON Schema management for input/output definitions
-
-## To research
-
-- Schema.org ld+json // Spatie https://github.com/spatie/schema-org // https://developers.google.com/search/docs/appearance/structured-data?hl=pl
-- OpenAPI Schema
-- nette/schema https://github.com/nette/schema
-- https://flow-php.com/
-- Queue-based load leveling
-- Throttling
-- Circuit breaker
-- Producer-consumer / queue-worker
-- Rate limiting
-- Retry on service failure
-- Backpressure
-- Batch stage chain
-- Request aggregator
-- Rolling poller window
-- Sparse task scheduler
-- Marker and sweeper
-- Actor model
-
-
-
-
-# Other
-
-## Gaps or issues in docs or code
-
-## Design decisions to revisit
-
-## Example ideas
-
-Examples to demonstrate use cases.
-
-## Test coverage
-
-Catch up with the latest additions.
-
-
-
-# Done
-
-> NOTE: Move notes here
-
-## Configuration
-
-- Examples how to override default configuration
-
-## Evals
-
-- Simplify contracts - currently 5 (!) contracts for observations
-- Add input, output, etc. tokens default metrics
-
+# NOTES
+
+## High priority
+
+- Evals / eval framework
+  * execution level correctness metric
+- Add 'Output' section to each example, generate it and include in docs, so reader can see what they can expect
+- Logging via PSR-3
+- Schema abstraction layer - decouple names and descriptions from the model
+- Prompt optimization via TextGrad
+- Agents
+- Indexing to vector DB
+- CLI app
+- Fast/simple REST API server - compatible with OpenAI? 
+- Revise examples debugging - not sure if it works as expected (what does it demonstrate?)
+- Multiple tools with tool selection
+- Parallel tool calls
+- Export configuration to user folder / use external configuration
+- Full control over generated prompt (access to Script object processing)
+- Gemini OpenAI-style API
+
+## Low priority
+
+- Batch API support (Gemini, OpenAI, Anthropic)
+- Gemini context caching
+- Generate unstructured, then format to structured - to improve reasoning
+
+
+# BACKLOG
+
+## Addon: Modules
+
+- Modules: Add module observability via events - currently no access to this info
+
+
+## Refactorings
+
+- More modular design - serialization, validation, transformation should be a configurable pipeline
+- Rework Events so they have toArray() method, make __toString() use it
+- Role - should be enum, not string?
+- String >> Array >> Class - for example: prompts (they should be classes)
+- Prompt - should be a class, not a string; prompt translates to Section/Messages; alt name: Instruction(s)
+
+## Infrastructure
+
+- Async mode
+- Fix response caching
+- Finish logging support
+- Document how to inject custom HTTP client
+- Better error messages
+- Documentation for logging
+- PSR-14 events - finish, demo how to plug custom dispatcher
+- PSR-11 container - finish, demo how to plug custom container
+- PSR-3 logger - finish, demo how to plug custom logger
+
+## Other
+
+- Sequences - validate only individual items, reject ONLY the invalid; allows progressive extraction in multiple stages
+- Use LLM to generate Example based on the class - just render object to JSON + add schema as an explanation - this should give the model enough info to come up with something that makes sense
+- Extract APIClient to a separate package?
+- Foreign client support - OpenAI PHP, Amazon Bedrock SDK, etc. - via some kind of adapters?
+- example of integration with Laravel/Livewire
+- Read rate limits from API responses
+- Rate limited API calls - wait for the limit to reset
+
+
+
+# Partially done
+
+- Better API for image / audio inputs
+- Make system, prompt, script etc. available for configuration by user
+- How to track API rate limits across multiple requests / parallel executions
+- Make using DocBlocks optional - it may not always to be desired to pass this info to LLM
+- Add super detailed tests of Module core functionality - esp. around input/output mappings
+- Validators / Deserializers / Transformers - chain of objects, not a single object
+- API Client: Clean up predefined models, prices, etc.
+
+
+
+
+# Brain dump
+
+- Streaming JSON parser - https://github.com/kelunik/streaming-json
+- Retry - https://github.com/kelunik/retry
+- Rate limiter - https://github.com/kelunik/rate-limit
+- Check for broken links - https://github.com/kelunik/link-check
+- Async flow execution - https://github.com/darkwood-com/flow
+- Models to implement JsonSerializable: https://www.php.net/manual/en/jsonserializable.jsonserialize.php / https://www.sitepoint.com/use-jsonserializable-interface/
+- Support JsonException for serialization / deserialization errors - https://www.php.net/manual/en/class.jsonexception.php
+- ValueError - https://www.php.net/manual/en/class.valueerror.php
+- Task runner - https://robo.li/
+- Hub >> Laravel Zero - https://laravel-zero.com/
+- DSPy next steps: eval, optimize, compile
+- Add more modules: XoT, RAG, ReAct, etc.
+- RAG - how to handle multiple VDB providers?
+- Test validation in modules - provide an example
+- Parallel execution of modules (e.g. a la Laravel jobs?)
+- Moderation endpoint support
+- Git/GitHub integration module to allow easy automation
+- Data mapping module(s) for easier data transformations
+- How to handle dynamic module graph definition + visualization
+- JSON Schema management for input/output definitions
+
+## To research
+
+- Schema.org ld+json // Spatie https://github.com/spatie/schema-org // https://developers.google.com/search/docs/appearance/structured-data?hl=pl
+- OpenAPI Schema
+- nette/schema https://github.com/nette/schema
+- https://flow-php.com/
+- Queue-based load leveling
+- Throttling
+- Circuit breaker
+- Producer-consumer / queue-worker
+- Rate limiting
+- Retry on service failure
+- Backpressure
+- Batch stage chain
+- Request aggregator
+- Rolling poller window
+- Sparse task scheduler
+- Marker and sweeper
+- Actor model
+
+
+
+
+# Other
+
+## Gaps or issues in docs or code
+
+## Design decisions to revisit
+
+## Example ideas
+
+Examples to demonstrate use cases.
+
+## Test coverage
+
+Catch up with the latest additions.
+
+
+
+# Done
+
+> NOTE: Move notes here
+
+## Configuration
+
+- Examples how to override default configuration
+
+## Evals
+
+- Simplify contracts - currently 5 (!) contracts for observations
+- Add input, output, etc. tokens default metrics
+
diff --git a/prompts/misc/cia.twig b/prompts/misc/cia.twig