Introduces SharedBackend Protocol for cross-process state sharing:
- InMemoryBackend: default single-process storage
- ManagerBackend: multiprocessing.Manager for local multi-process
- RedisBackend: distributed deployments with TTL auto-GC
Adds ProcessPoolExecutor support for CPU-bound handlers:
- worker.py: worker process entry point
- stream_pump.py: cpu_bound handler dispatch
- Config: backend and process_pool sections in organism.yaml
ContextBuffer and ThreadRegistry now accept optional backend
parameter while maintaining full backward compatibility.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
These modules are now proprietary and live in the Nextra SaaS product.
xml-pipeline remains the OSS core with:
- Message pump and pipeline steps
- Handler contract and responses
- LLM router abstraction
- Native tools
- Config loading
- Memory/context buffer
Removed:
- xml_pipeline/console/ → nextra/console/
- xml_pipeline/auth/ → nextra/auth/
- xml_pipeline/server/ → nextra/server/
- Legacy files: agentserver.py, main.py, xml_listener.py
The simple console example remains in examples/console/.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Use __file__-based path resolution for envelope.xsd so the schema
loads correctly when xml-pipeline is installed via pip.
Also:
- Add build artifacts to .gitignore
- Bump version to 0.3.1
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Remove unused Color type import from termcolor.termcolor which doesn't
exist in newer termcolor versions. Change type hints from Color to str.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
README.md:
- Rebrand from AgentServer to xml-pipeline
- Library-focused introduction with pip install
- Quick start guide with code examples
- Console example documentation
- Concise feature overview
pyproject.toml:
- Update authors to "xml-pipeline contributors"
- Update URLs to xml-pipeline.org
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
OSS restructuring for open-core model:
- Rename package from agentserver/ to xml_pipeline/
- Update all imports (44 Python files, 31 docs/configs)
- Update pyproject.toml for OSS distribution (v0.3.0)
- Move prompt_toolkit from core to optional [console] extra
- Remove auth/server/lsp from core optional deps (-> Nextra)
New console example in examples/console/:
- Self-contained demo with handlers and config
- Uses prompt_toolkit (optional, falls back to input())
- No password auth, no TUI, no LSP — just the basics
- Shows how to use xml-pipeline as a library
Import changes:
- from agentserver.* -> from xml_pipeline.*
- CLI entry points updated: xml_pipeline.cli:main
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- SystemPipeline: Entry point for console/webhook/API messages
- TextInput/TextOutput: Generic primitives for human text I/O
- Server: WebSocket "send" command routes through SystemPipeline
- Console: @target message now injects into pipeline
Flow: Console → WebSocket → SystemPipeline → XML envelope → pump
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- auth/users.py: User store with Argon2id password hashing
- auth/sessions.py: Token-based session management with expiry
- server/app.py: aiohttp server with auth middleware and WebSocket
- console/client.py: SSH-style login console client
Server endpoints: /auth/login, /auth/logout, /auth/me, /health, /ws
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Lua proposal covers:
- Sandboxed Lua handlers for SaaS custom logic
- XML <-> Lua table bridging with XSD validation
- Security sandbox (restricted stdlib, resource limits)
- Schema DSL alternative to raw XSD
- Hot reload mechanism
API spec updated with:
- Dedicated /ws/messages endpoint for message log streaming
- Filter by agent/thread/payload type
- Example for reconstructing agent buffers client-side
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Covers authentication (TOTP, WebAuthn, mTLS, OAuth), authorization
(RBAC, resource permissions), transport security, secrets management,
audit logging, prompt injection defense, and rate limiting.
Includes open questions and implementation checklist.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Defines the GUI-agnostic API for monitoring and controlling the organism:
- REST endpoints for queries and control
- WebSocket for real-time push events
- Data models for agents, messages, threads
- Example usage for 3D office GUI
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- is_at_bottom() now checks if on last 3 lines (not exact end)
- Page Down snaps to bottom when within 3 lines of end
- Added scroll_to_bottom() helper method
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Spacer only shows when output is at bottom (disappears when scrolled up)
- New content appends silently when user has scrolled up (preserves position)
- is_at_bottom() helper to check scroll state
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
auto_up/auto_down were wrong - use history_backward/history_forward
for proper command history navigation.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Up/Down arrows now navigate command history via auto_up/auto_down
- Added blank line spacer between output and separator for clarity
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Standard CLI behavior:
- Up/Down = command history (always)
- Page Up/Down = scroll output (no focus needed)
- Ctrl+Home/End = jump to top/bottom
- Mouse/scrollbar works too
Removed Tab focus switching - not intuitive for CLI users.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Use Condition wrapper with try/except to handle case where
self.app doesn't exist yet during key binding setup.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Up/Down move cursor one line (scrolls view)
- Page Up/Down move 20 lines at a time
- Home/End jump to top/bottom
- Keys only active when output buffer is focused (Tab to switch)
- Updated help text with scroll shortcuts
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
FormattedTextControl doesn't support scrolling - always renders from top.
Switch to Buffer + BufferControl which has proper scroll support:
- Output buffer uses read-only Buffer with cursor at end
- Tab/Shift-Tab to switch focus between output and input
- Arrow keys scroll when output is focused
- Scrollbar margin shows position
- Auto-scroll on new content (cursor stays at end)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Window.vertical_scroll is read-only in prompt_toolkit.
Use custom scroll_offset to track position and slice output lines.
- scroll_offset=0 shows newest content (auto-scroll)
- Page Up increases offset (scroll into history)
- Page Down decreases offset (towards newest)
- Home/End jump to top/bottom
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add Page Up/Down, Home/End key bindings for manual scrolling
- Track user_scrolled flag to preserve position when user scrolls
- Remove visible_lines limit to show full history
- Auto-scroll to bottom on new content (unless user scrolled up)
- Update help text with scroll shortcuts
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Set vertical_scroll=999999 to always show newest content
- Update scroll position on every invalidate
- Reduced visible_lines to 20 for safety
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Use HSplit with filler + output to push content to bottom
- Output window uses dont_extend_height so it only takes needed space
- Empty filler expands to fill remaining space above
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Instead of trying to scroll the window, just display the most
recent lines that fit on screen. Full history is kept but only
the last 50 lines are rendered.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Create separate console_registry module to avoid __main__ vs module import issue
- Handler now correctly finds TUI console instance
- Response output should now appear in TUI output area
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Remove patch_stdout (was buffering until exit)
- Add 100ms refresh loop that periodically invalidates the app
- Ensures output buffer changes are displayed in real-time
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add patch_stdout to capture external prints in TUI mode
- Any stdout from handlers/pump appears in output area, not prompt
- Simplify response handler logic
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add thread-safe invalidation for cross-task updates
- Fix response handler to properly route to TUI output
- Add debug output if console hook fails
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
New console features:
- Scrolling output area above separator line
- Status bar with organism name
- Input area at bottom
- Color-coded messages by source
- Command history with up/down arrows
- Graceful fallback to simple mode on unsupported terminals
Usage:
python run_organism.py # TUI mode (or fallback)
python run_organism.py --simple # Force simple mode
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Import `complete` not `generate` from agentserver.llm
- Pass `model` parameter required by router
- Return response.content (string) not response object
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Platform architecture for trusted orchestration:
- PromptRegistry: immutable system prompts per agent, loaded at bootstrap
- platform.complete(): assembles LLM calls (prompt + history + user msg)
- Handlers use platform API, cannot see/modify prompts
- organism.yaml now supports prompt field per listener
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Documents the trust model for context and prompt management:
- Platform as trusted orchestration layer
- Context buffer: pure message history, keyed by UUID, auto-GC
- Prompt registry: immutable, invisible to agents
- LLM call assembly: platform controls, agents request
- Handler sandbox: what agents can/cannot do
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Shows last 20 messages from a specific thread or last 5 per thread
when using /monitor *. Color-coded by direction (console=green,
responses=cyan).
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- SecureConsole with Argon2id password hashing
- Password stored in ~/.xml-pipeline/console.key
- Protected commands require password re-entry
- Attach/detach model with idle timeout
- Fallback to simple input for incompatible terminals
- @listener message injection into pump
- Boot handler no longer sends to old console
- Response handler prints to stdout with prompt refresh
Dependencies: argon2-cffi, prompt_toolkit
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
