abstractcode 0.2.0__py3-none-any.whl → 0.3.0__py3-none-any.whl

abstractcode-0.3.0.dist-info/METADATA

@@ -0,0 +1,270 @@
+Metadata-Version: 2.4
+Name: abstractcode
+Version: 0.3.0
+Summary: A clean terminal CLI for multi-agent agentic coding
+Author-email: Laurent-Philippe Albou <contact@abstractcore.ai>
+Maintainer-email: Laurent-Philippe Albou <contact@abstractcore.ai>
+License: MIT
+Project-URL: Homepage, https://abstractcore.ai
+Project-URL: Documentation, https://abstractcore.ai
+Project-URL: Repository, https://github.com/lpalbou/abstractcode
+Project-URL: Bug Tracker, https://github.com/lpalbou/abstractcode/issues
+Keywords: ai,llm,cli,coding,agent,multi-agent,agentic,terminal,abstractcore,abstractruntime,abstractagent
+Classifier: Development Status :: 2 - Pre-Alpha
+Classifier: Intended Audience :: Developers
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Topic :: Software Development :: Libraries :: Python Modules
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Requires-Python: >=3.10
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Requires-Dist: abstractagent>=0.2.0
+Requires-Dist: abstractruntime>=0.2.0
+Requires-Dist: abstractcore[tools]>=2.6.8
+Requires-Dist: ddgs<10.0.0,>=9.10.0
+Requires-Dist: prompt_toolkit>=3.0.0
+Provides-Extra: dev
+Requires-Dist: pytest>=7.0.0; extra == "dev"
+Requires-Dist: pytest-cov>=4.0.0; extra == "dev"
+Requires-Dist: black>=23.0.0; extra == "dev"
+Requires-Dist: ruff>=0.1.0; extra == "dev"
+Provides-Extra: flow
+Requires-Dist: abstractflow>=0.1.0; extra == "flow"
+Dynamic: license-file
+
+# AbstractCode
+
+Terminal TUI for the AbstractFramework:
+- run **agents** (ReAct / CodeAct)
+- run **workflows** authored in AbstractFlow (VisualFlow JSON)
+- keep everything **durable** via AbstractRuntime (runs, ledger, artifacts)
+
+## Install
+
+```bash
+pip install abstractcode
+```
+
+To run AbstractFlow workflows from AbstractCode:
+
+```bash
+pip install "abstractcode[flow]"
+```
+
+## Start the TUI (Agents)
+
+```bash
+abstractcode --provider lmstudio --model qwen/qwen3-next-80b
+```
+
+Interaction model:
+- commands are slash-prefixed (`/help`)
+- any non-command line starts a task (same as `/task ...`)
+
+### Useful commands
+- `/status` (run status)
+- `/auto-accept on|off` (tool approvals)
+- `/max-tokens N` (or `-1` auto-detect)
+- `/compact [light|standard|heavy] [--preserve N] [focus...]` (durable compaction)
+- `/spans`, `/expand <span> [--show] [--into-context]` (provenance recall)
+- `/recall [--since ISO] [--until ISO] [--tag k=v] [--q text] [--into-context]`
+- `/snapshot save|load|list`
+
+### Persistence
+By default, AbstractCode uses `~/.abstractcode/state.json` and stores the durable data in `~/.abstractcode/state.d/`.
+- disable persistence: `abstractcode --no-state`
+- override path: `ABSTRACTCODE_STATE_FILE=... abstractcode`
+
+## Workflow Agents (VisualFlow as `--agent`)
+
+AbstractCode can run an AbstractFlow VisualFlow workflow *as an agent* (instead of using the built-in `react|codeact|memact` agents).
+
+### Requirements (`abstractcode.agent.v1`)
+- The workflow JSON must declare: `interfaces: ["abstractcode.agent.v1"]`
+- The workflow must expose these pins:
+  - `On Flow Start`: output pins:
+    - `request` (type `string`)
+    - `provider` (type `provider`)
+    - `model` (type `model`)
+    - `tools` (type `tools`)
+  - `On Flow End`: input pin `response` (type `string`)
+
+Recommended (optional) pins:
+- `On Flow Start`: `context` (object), `max_iterations` (number)
+- `On Flow End`: `meta` (object), `scratchpad` (object), `raw_result` (object)
+
+### Authoring in the AbstractFlow visual editor
+- **Mark the interface**: click `📂 Load` → select the workflow → in the right preview panel find **Interfaces** → click the ✏️ icon → enable **AbstractCode Agent (v1)** → **Save Interfaces**
+- **Pins are scaffolded automatically**: when the interface is enabled, AbstractFlow will ensure `On Flow Start` and `On Flow End` have the required pins. You can still add/remove optional pins as needed.
+
+Tip: an example workflow is shipped at `abstractflow/web/flows/acagent01.json` (implements the interface).
+
+### What is `meta` and how do I use it?
+`On Flow End.meta` is an **optional JSON object** for host-facing metadata (usage, trace ids, warnings, raw provider info, etc.).
+
+It is intentionally **not strictly validated** today (the host treats it as opaque JSON). To make workflows portable and predictable across hosts, we recommend using a small “envelope” shape:
+
+```json
+{
+  "schema": "abstractcode.agent.v1.meta",
+  "version": 1,
+  "provider": "lmstudio",
+  "model": "qwen/qwen3-next-80b",
+  "usage": { "input_tokens": 123, "output_tokens": 456 },
+  "trace": { "trace_id": "..." },
+  "warnings": ["..."],
+  "debug": {}
+}
+```
+
+Hosts should treat unknown fields as allowed and ignore what they don’t understand (forward-compatible).
+
+Typical ways to produce it inside a workflow:
+- Wire `LLM Call.result` (object) → `On Flow End.meta`
+- Wire `Agent.result` (object) → `On Flow End.meta`
+- Or build your own object and wire it into `meta`
+
+When present, AbstractCode attaches it to the assistant message metadata as `workflow_meta`.
+
+### Workflow-driven status updates (footer / live UX)
+Inside a workflow, you can update AbstractCode’s footer status text by emitting the reserved event:
+- Add an **Emit Event** node
+- Set **name** to `abstractcode.status`
+- Set **payload** to:
+  - a string (e.g. `"Enrich Query..."`), or
+  - an object like `{ "text": "Enrich Query...", "duration": -1 }`
+
+`duration` is seconds:
+- default: `-1` (sticky)
+- if `> 0`: auto-clears after the timeout unless superseded by a newer status
+
+Example workflow: `abstractflow/web/flows/acagent_status_demo.json` (3 status updates, each separated by a 2s Delay).
+
+### Workflow-driven UI events (messages + tool UX)
+Workflows can also emit additional reserved events for host UX:
+- `abstractcode.message`: show a message/notification (payload is a string or `{text, level?, title?}`)
+- `abstractcode.tool_execution`: render a tool-call block (payload is a tool call object or a list)
+  - recommended shape: `{name, arguments, call_id?}`
+- `abstractcode.tool_result`: render a tool-result block (payload is a tool result object or a list)
+  - recommended shape: `{name, call_id?, success?, output?, error?}`
+
+Full contract (recommended for integrators): `docs/ui_events.md`
+
+Example workflows:
+- `abstractflow/web/flows/acagent_message_demo.json`
+- `abstractflow/web/flows/acagent_tool_events_demo.json`
+
+### Durable ask+wait (prompt the user and resume)
+For workflows that need human input, you can:
+- use the **Ask User** node (WAIT_USER), or
+- use **Wait Event** with a `prompt` field (WAIT_EVENT), which is also durable and network-friendly.
+
+Example workflow: `abstractflow/web/flows/acagent_ask_demo.json`
+
+### Run
+Use `--agent` with a workflow id/name (from the flows directory) or a direct JSON path:
+
+```bash
+abstractcode --agent acagent01
+abstractcode --agent abstractflow/web/flows/acagent01.json
+
+# If your flows are stored elsewhere:
+ABSTRACTFLOW_FLOWS_DIR=/path/to/flows abstractcode --agent my_flow_id
+```
+
+## Run Workflows (AbstractFlow VisualFlow)
+
+### From the CLI
+
+```bash
+abstractcode flow run <flow_id_or_path> [inputs...]
+```
+
+Inputs can be passed as flags (no JSON typing required):
+
+```bash
+abstractcode flow run abstractflow/web/flows/4e2f2329.json --query "who are you?"
+
+abstractcode flow run abstractflow/web/flows/b3a9d7c1.json \
+  --query "who are you?" \
+  --max_web_search 15 \
+  --max_fetch_url 50 \
+  --follow_up_questions true
+```
+
+Other input options:
+
+```bash
+abstractcode flow run deep-research-pro --input-json-file params.json
+abstractcode flow run deep-research-pro --param max_web_search=15 --param follow_up_questions=true
+```
+
+Tool approvals:
+- approval-gated by default (type `a` once to approve all remaining calls for that run)
+- auto-approve (unsafe): `--accept-tools` (alias `--auto-approve`)
+
+Controls:
+
+```bash
+abstractcode flow resume
+abstractcode flow pause
+abstractcode flow resume-run
+abstractcode flow cancel
+```
+
+Run discovery and event injection:
+
+```bash
+abstractcode flow runs
+abstractcode flow attach <run_id>
+abstractcode flow emit --name my_event --scope session --payload-json '{"k":"v"}'
+```
+
+Flow state:
+- default: `~/.abstractcode/flow_state.json` (stores in `~/.abstractcode/flow_state.d/`)
+- override with `ABSTRACTCODE_FLOW_STATE_FILE=...`
+- flow discovery default dir: `ABSTRACTFLOW_FLOWS_DIR=...`
+
+### From inside the TUI (keeps outputs in context)
+
+```text
+/flow run deep-research-pro --query "..." --max_web_search 10 --follow_up_questions true
+```
+
+`ANSWER_USER` outputs from the workflow are appended into the current conversation’s active context so you can continue the dialogue naturally.
+
+## Development (Monorepo)
+
+```bash
+pip install -e ./abstractcore -e ./abstractruntime -e ./abstractagent -e ./abstractflow -e ./abstractcode
+abstractcode --help
+```
+
+## Environment variables
+- `ABSTRACTCODE_STATE_FILE`
+- `ABSTRACTCODE_FLOW_STATE_FILE`
+- `ABSTRACTFLOW_FLOWS_DIR`
+
+## Default Tools
+
+AbstractCode provides a curated set of 10 tools for coding tasks (ReAct agent):
+
+| Tool | Description |
+|------|-------------|
+| `list_files` | Find and list files using glob patterns (case-insensitive) |
+| `search_files` | Search for text patterns inside files using regex |
+| `analyze_code` | Outline a Python/JS file (imports/classes/functions + line ranges) |
+| `read_file` | Read file contents with optional line range |
+| `write_file` | Write content to files, creating directories as needed |
+| `edit_file` | Edit files by replacing text patterns (supports regex, line ranges, preview mode) |
+| `execute_command` | Execute shell commands with security controls |
+| `web_search` | Search the web via DuckDuckGo (no API key required) |
+| `fetch_url` | Fetch a URL and return text/metadata (best-effort parsing) |
+| `self_improve` | Log improvement suggestions for later review |
+
+When running `--agent codeact`, AbstractCode exposes `execute_python` instead of the ReAct toolset.

abstractcode-0.3.0.dist-info/RECORD

@@ -0,0 +1,17 @@
+abstractcode/__init__.py,sha256=pHrawn8KtpITMmArbfOQBm1WrGKL7qMAr0cu81ianDU,658
+abstractcode/cli.py,sha256=2DS9eI9VgFcxlE78Pom_Q5OsiMQxA8PKOGyVXDK3qnQ,14797
+abstractcode/flow_cli.py,sha256=CQq1g4rFylSW7y8ZZeQYmfroAXG1dDtmuOF8tc1BF2Y,49235
+abstractcode/fullscreen_ui.py,sha256=3pRz098lbTbmPH14w2zKVPXUeTFVTtCyRSNs2h3w_yc,59898
+abstractcode/input_handler.py,sha256=W2ImW4Yr3HMMD5SNPZOZJendVL-CRdYR5idXTWroxAY,2240
+abstractcode/py.typed,sha256=AbpHGcgLb-kRsJGnwFEktk7uzpZOCcBY74-YBdrKVGs,1
+abstractcode/react_shell.py,sha256=0Dk9KrIsvRNC4ufRFC6WQwVjY2TzS73gUUYRNKP_NeE,285565
+abstractcode/recall.py,sha256=xu-xz8-kvkmAW6pKoAIcwox5jHWMBfBKExsA8fgwg_A,13040
+abstractcode/remember.py,sha256=cmJlw-nR0OvK7tkN5B0sZSMaYlECl_tAyO34c-IL0Uo,6158
+abstractcode/terminal_markdown.py,sha256=mDR-70S-P6dv4JefB6OgmjQRw9tFtP7RNkW9ZHw8NLE,5449
+abstractcode/workflow_agent.py,sha256=belBm7TyyO1zu57H70hklB3mos3BR43FpvRdbJDgk_A,35323
+abstractcode-0.3.0.dist-info/licenses/LICENSE,sha256=f8z-PbFM_xH0eqLsha8_nEg1jbm1eVtC2NNjtXhtuCo,1080
+abstractcode-0.3.0.dist-info/METADATA,sha256=sbuUa8yTLaJHgchguiqH5ee0hIhnrKmw3nZ63MBttag,10179
+abstractcode-0.3.0.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
+abstractcode-0.3.0.dist-info/entry_points.txt,sha256=M4AOKb4h2vKWy5OMg-czeJE2aXunVKlljUPN8BUtwoE,51
+abstractcode-0.3.0.dist-info/top_level.txt,sha256=n9Hfv6cnmL55HoxXsA3V1YLbQxYAr58-UMLbdr-FCNw,13
+abstractcode-0.3.0.dist-info/RECORD,,

abstractcode-0.2.0.dist-info/METADATA

@@ -1,160 +0,0 @@
-Metadata-Version: 2.4
-Name: abstractcode
-Version: 0.2.0
-Summary: A clean terminal CLI for multi-agent agentic coding
-Author-email: Laurent-Philippe Albou <contact@abstractcore.ai>
-Maintainer-email: Laurent-Philippe Albou <contact@abstractcore.ai>
-License: MIT
-Project-URL: Homepage, https://abstractcore.ai
-Project-URL: Documentation, https://abstractcore.ai
-Project-URL: Repository, https://github.com/lpalbou/abstractcode
-Project-URL: Bug Tracker, https://github.com/lpalbou/abstractcode/issues
-Keywords: ai,llm,cli,coding,agent,multi-agent,agentic,terminal,abstractcore,abstractruntime,abstractagent
-Classifier: Development Status :: 2 - Pre-Alpha
-Classifier: Intended Audience :: Developers
-Classifier: License :: OSI Approved :: MIT License
-Classifier: Operating System :: OS Independent
-Classifier: Programming Language :: Python :: 3
-Classifier: Programming Language :: Python :: 3.8
-Classifier: Programming Language :: Python :: 3.9
-Classifier: Programming Language :: Python :: 3.10
-Classifier: Programming Language :: Python :: 3.11
-Classifier: Programming Language :: Python :: 3.12
-Classifier: Topic :: Software Development :: Libraries :: Python Modules
-Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
-Requires-Python: >=3.8
-Description-Content-Type: text/markdown
-License-File: LICENSE
-Requires-Dist: prompt_toolkit>=3.0.0
-Provides-Extra: dev
-Requires-Dist: pytest>=7.0.0; extra == "dev"
-Requires-Dist: pytest-cov>=4.0.0; extra == "dev"
-Requires-Dist: black>=23.0.0; extra == "dev"
-Requires-Dist: ruff>=0.1.0; extra == "dev"
-Dynamic: license-file
-
-# AbstractCode
-
-**A clean terminal CLI for multi-agent agentic coding**
-
-[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://opensource.org/licenses/MIT)
-[![Python 3.8+](https://img.shields.io/badge/python-3.8+-blue.svg)](https://www.python.org/downloads/)
-
----
-
-## Status
-
-AbstractCode is under active development. A minimal interactive shell exists to support manual testing of AbstractAgent workflows.
-
-Note: the PyPI release may lag behind the monorepo. For the latest development version, install from source.
-
-## What is AbstractCode?
-
-AbstractCode is a clean terminal CLI for multi-agent agentic coding, similar to Claude Code, Codex, and Gemini CLI. It leverages the powerful Abstract Framework ecosystem to provide seamless AI-powered coding assistance directly in your terminal.
-
-## The Abstract Framework
-
-AbstractCode is built on top of the Abstract Framework, a comprehensive suite of tools for AI-powered development:
-
-- **[AbstractCore](https://github.com/lpalbou/abstractcore)** - Unified interface for multiple LLM providers
-- **[AbstractRuntime](https://github.com/lpalbou/abstractruntime)** - Runtime environment for AI agents
-- **[AbstractAgent](https://github.com/lpalbou/abstractagent)** - Multi-agent orchestration and coordination
-
-## Features (Coming Soon)
-
-- 🤖 **Multi-Agent Coding** - Coordinate multiple AI agents for complex coding tasks
-- 🔌 **Provider Agnostic** - Works with OpenAI, Anthropic, Ollama, and more
-- 💻 **Terminal Native** - Clean CLI interface for seamless workflow integration
-- 🎯 **Context Aware** - Understands your codebase and project structure
-- 🔄 **Iterative Development** - Collaborative coding with AI assistance
-- 🌐 **Offline Capable** - Works with local models via Ollama
-
-## Installation
-
-```bash
-pip install abstractcode
-```
-
-## Quick Start
-
-```bash
-# Show options
-abstractcode --help
-
-# Durable resume is enabled by default (state file: ~/.abstractcode/state.json)
-# Override with:
-ABSTRACTCODE_STATE_FILE=.abstractcode.state.json abstractcode
-
-# Or disable persistence (in-memory only; cannot resume after quitting)
-abstractcode --no-state
-
-# Auto-approve tool calls (unsafe; bypasses interactive approvals)
-abstractcode --auto-approve
-
-# Limit agent iterations per task (default: 20)
-abstractcode --max-iterations 25
-```
-
-Notes:
-- Run resume state is stored next to the state file in `*.d/`.
-- Conversation history is stored in the run state (`RunState.vars["context"]["messages"]`) inside `*.d/`, and AbstractCode keeps the state file pointing at the most recent run so restarts can reload context.
-- In the interactive shell, commands are slash-prefixed (e.g. `/help`, `/status`, `/history`, `/task ...`).
-
-## Development (Monorepo)
-
-From the monorepo root:
-
-```bash
-pip install -e ./abstractcore -e ./abstractruntime -e ./abstractagent -e ./abstractcode
-abstractcode --help
-```
-
-## Requirements
-
-- Python 3.8 or higher
-- AbstractCore
-- AbstractRuntime
-- AbstractAgent
-
-## Documentation
-
-Full documentation will be available at [abstractcore.ai](https://abstractcore.ai)
-
-## Development Status
-
-This project is in early development. Stay tuned for updates!
-
-## Contributing
-
-Contributions are welcome! Please check back soon for contribution guidelines.
-
-## Contact
-
-**Maintainer:** Laurent-Philippe Albou  
-📧 Email: contact@abstractcore.ai  
-🌐 Website: [abstractcore.ai](https://abstractcore.ai)
-
-## License
-
-MIT License - see LICENSE file for details.
-
----
-
-**AbstractCode** - Multi-agent agentic coding in your terminal, powered by the Abstract Framework.
-
-## Default Tools
-
-AbstractCode provides a curated set of 8 tools for coding tasks:
-
-| Tool | Description |
-|------|-------------|
-| `list_files` | Find and list files using glob patterns (case-insensitive) |
-| `search_files` | Search for text patterns inside files using regex |
-| `read_file` | Read file contents with optional line range |
-| `write_file` | Write content to files, creating directories as needed |
-| `edit_file` | Edit files by replacing text patterns (supports regex, line ranges, preview mode) |
-| `execute_command` | Execute shell commands with security controls |
-| `web_search` | Search the web via DuckDuckGo (no API key required) |
-| `self_improve` | Log improvement suggestions for later review |
-
-Additional tools are available via AbstractAgent for specialized use cases (execute_python, fetch_url).

abstractcode-0.2.0.dist-info/RECORD

@@ -1,11 +0,0 @@
-abstractcode/__init__.py,sha256=pHrawn8KtpITMmArbfOQBm1WrGKL7qMAr0cu81ianDU,658
-abstractcode/cli.py,sha256=oW9S7eZlsIFP2yF9TtlTi6vaUF120uqanLnsntoSSL0,3438
-abstractcode/fullscreen_ui.py,sha256=-7xOhz657TXjZvcQeCzJ3RfzyaEuTzXI6e5uge7U0V4,23951
-abstractcode/input_handler.py,sha256=W2ImW4Yr3HMMD5SNPZOZJendVL-CRdYR5idXTWroxAY,2240
-abstractcode/react_shell.py,sha256=fOLiTJje3IeJ0LjAyye_vsN6m7CLNx9R6skDGXN2258,50699
-abstractcode-0.2.0.dist-info/licenses/LICENSE,sha256=f8z-PbFM_xH0eqLsha8_nEg1jbm1eVtC2NNjtXhtuCo,1080
-abstractcode-0.2.0.dist-info/METADATA,sha256=38oM_FIIMKz2jpCxKOg4fA_vnLv8OD3v2zDUtmC_TME,5985
-abstractcode-0.2.0.dist-info/WHEEL,sha256=_zCd3N1l69ArxyTb8rzEoP9TpbYXkqRFSNOD5OuxnTs,91
-abstractcode-0.2.0.dist-info/entry_points.txt,sha256=M4AOKb4h2vKWy5OMg-czeJE2aXunVKlljUPN8BUtwoE,51
-abstractcode-0.2.0.dist-info/top_level.txt,sha256=n9Hfv6cnmL55HoxXsA3V1YLbQxYAr58-UMLbdr-FCNw,13
-abstractcode-0.2.0.dist-info/RECORD,,