Releases: AndrewClements84/PromptStream.AI
v0.8.5 Costing Hotfix
Hotfix to correct costing calculations
v0.8.3 — Accurate Cost Estimation + CLI Stability Update
🏷️ Release v0.8.3 — Accurate Cost Estimation + CLI Stability Update
🚀 Highlights
- Fixed Cost Estimation:
CLI now accurately calculates token costs using embedded TokenFlow.AI registry data (supports both legacy and modern pricing fields). - Improved Model Registry Integration:
Automatically loads model metadata from TokenFlow.AI without requiring external configuration. - CLI Packaging Polished:
Updated.csprojfor clean NuGet tool packaging (PackAsTool=true), ensuring global installs (dotnet tool install -g promptstream) work seamlessly. - Internal Enhancements:
- Simplified and documented
PromptStreamService.AnalyzePrompt() - Added embedded JSON deserialization shim for cross-version registry compatibility
- Verified cost calculations in both Debug and global tool contexts
- Simplified and documented
🧩 Components Updated
PromptStream.AIPromptStream.AI.CLI- Integration with
TokenFlow.AI(no republish required)
🧪 Verification
promptstream analyze --template "Explain neural networks simply." --model gpt-4oExpected output:
📊 Prompt Analysis
────────────────────────────────────────
Model: gpt-4o
Prompt Tokens: 4
Completion Tokens: 0
Total Tokens: 4
Estimated Cost: $0.010000
────────────────────────────────────────
🧠 Notes
- No API surface changes — all interfaces remain backward-compatible.
- Version
0.8.3is safe to publish as both a NuGet library and a .NET global tool.
v0.8.1 CLI V2
v0.8.1
- Added new
analyzecommand for token and cost analysis - Updated documentation (
cli-analyze.md) - Minor CLI help and formatting improvements
PromptStream.AI v0.8.0 — Unified with Flow.AI.Core
PromptStream.AI v0.8.0 introduces full integration with Flow.AI.Core v0.2.0, unifying all shared models and interfaces across the Flow ecosystem. This release removes redundant local types, refactors internal services and tests to use the shared DTOs, and maintains 100% test coverage and documentation parity. It marks the first fully Core-aligned version, paving the way for seamless interoperability with TokenFlow.AI, ReasonFlow.AI, and future Flow components.
🌟 Highlights
🧩 Integrated Flow.AI.Core v0.2.0 (shared models & interfaces)
🧱 Removed redundant local types and streamlined dependencies
✅ Refactored PromptBuilder, PromptContextManager, and all tests
📘 Added new documentation page: core-integration.md
💯 Maintained full test coverage and clean CI/CD workflows
PromptStream.AI v0.1.5 — Foundational Release (Core + TokenFlow Integration)
The first public release of PromptStream.AI, introducing a complete, modular toolkit for prompt composition, validation, and conversational context management in .NET.
🚀 Highlights
Core Library (PromptStream.AI)
- ✅ Token-aware prompt builder with variable interpolation
- ✅ Validation engine for completeness and token limits
- ✅ Context manager with merge, replay, summarization, and persistence (JSON)
- ✅ Token budgeting (
EstimateTokenUsage,TrimToTokenBudget) - ✅ Assistant reply tracking and multi-turn chat memory
- ✅ Seamless integration with TokenFlow.AI
- ✅ Comprehensive xUnit test coverage and Codecov integration
TokenFlow Integration (PromptStream.AI.Integration.TokenFlow)
- 💡 Provides
TokenFlowModelClientadapter implementingIModelClient - 🔗 Enables model-aware validation and token analysis from TokenFlow.AI
- 🧩 Ready for future Flow.AI.Core and ReasonFlow.AI interoperability
CLI (Developer Tooling)
- 💻 Added
PromptStream.AI.CLIproject for local dev tools - Commands:
build,validate,generate,context - Color-coded console output and context persistence
- Will be released later as a global tool (
dotnet tool install -g PromptStream.AI.CLI)
🧱 Dependencies
- .NET 8.0 or higher
- Flow.AI.Core
- TokenFlow.AI (for integration)
🧪 Testing
All components are covered by xUnit + Coverlet with full Codecov reporting.
🗺️ Next Steps
- Publish CLI as .NET Global Tool
- Add Token cost estimation
- OpenAI-style function call validation
- Flow.AI Studio visual editor integration
Part of the Flow.AI Ecosystem
© 2025 Andrew Clements