flexcoder 1.0.4

AI coding assistant TUI — multi-provider, session-aware

flexcoder

AI coding assistant TUI — alternative to opencode / claude code / codex.

Install

pip install flexcoder

File structure

flexcoder/
├── pyproject.toml
├── config.toml.default          template for ~/.flexcoder/config.toml
└── flexcoder/
    ├── main.py                  CLI entry point  (print("initialisation...") line 1)
    ├── config.py                TOML config loader/saver
    ├── providers.py             provider registry + verified fetch endpoints
    ├── sessions.py              session persistence → ~/.flexcoder/sessions/
    ├── ai.py                    AI backends (all 6 providers)
    ├── tools.py                 tool execution (create_file, shell, etc.)
    ├── system_prompt.py         dynamic system prompt (OS, shell, tree, tools)
    ├── art.py                   ASCII art
    ├── flexcoder.tcss           TUI stylesheet
    └── screens/
        ├── chat.py              main chat TUI
        ├── help.py              Ctrl+H help overlay
        ├── selector.py          Ctrl+P provider/model picker
        ├── provider_settings.py Ctrl+E API keys + Fetch Models
        ├── ai_settings.py       Ctrl+T temperature / max tokens / …
        └── session_browser.py   Ctrl+S session browser

Usage

flexcoder                           # start with saved config
flexcoder ollama gemma3:3b          # start with Ollama gemma3:3b
flexcoder continue                  # resume most recent session
flexcoder continue <session-id>     # resume specific session
flexcoder providers                 # provider settings TUI (standalone)
flexcoder settings                  # AI settings TUI (standalone)
flexcoder sessions                  # session browser TUI (standalone)

First-time setup

1. flexcoder → press Ctrl+E
2. Select provider, paste API key, click Save Key
3. Click Fetch Models (queries the provider's live API)
4. Click Done
5. Press Ctrl+P → select model → Confirm
6. Start chatting

Keyboard shortcuts

Key	Action
Ctrl+A	Toggle auto-approve (AUTO/MANUAL)
Ctrl+O	Toggle output (OUT/QUIET)
Ctrl+H	Help
Ctrl+N	New session
Ctrl+S	Session browser
Ctrl+P	Provider / model selector
Ctrl+E	Provider settings (API keys)
Ctrl+T	AI generation settings
Ctrl+L	Clear chat
Esc	Interrupt / cancel
Ctrl+C	Quit

Model fetch endpoints (verified March 2026)

Provider	Endpoint
Ollama	GET http://localhost:11434/api/tags
Claude	GET https://api.anthropic.com/v1/models (x-api-key header)
ChatGPT	GET https://api.openai.com/v1/models (Bearer token)
Gemini	GET https://generativelanguage.googleapis.com/v1beta/models?key=…
Mistral	GET https://api.mistral.ai/v1/models (Bearer token)
OpenRouter	GET https://openrouter.ai/api/v1/models (Bearer token)

AI tools available to the model

Tool	Syntax
Create file	<create_file=[path]>content</create_file>
Move file	<move_file source=[old] destination=[new] />
Read file	<read_file=[path] />
Create directory	<create_directory=[path] />
Search & replace	<search_replace=[file]><search>…</search><replace>…</replace></search_replace>
Shell command	<shell>command</shell>
Insert text	<insert=[file]><for>…</for><after>…</after><text>…</text></insert>

Session storage

Sessions live in ~/.flexcoder/sessions/<session-id>.json.
Resume any session: flexcoder continue 20250306-143022-a1b2c3