agentorch 2.0.1__tar.gz → 2.0.2__tar.gz

{agentorch-2.0.1 → agentorch-2.0.2}/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: agentorch
-Version: 2.0.1
+Version: 2.0.2
 Summary: DevTorch — AI reasoning capture, audit trail, and governance for agent-assisted development
 Author-email: Hemant Joshi <hemant@flotorch.ai>
 License: Apache-2.0
@@ -52,10 +52,10 @@ Requires-Dist: agentorch[cloud,mcp,proxy,wrapper]; extra == "all"
 
 **Local-first AI agent governance.** DevTorch gives multi-agent systems a shared on-disk audit trail, coordination vector, privacy controls, and LLM integration layer — all without a remote service.
 
-> **Latest release:** `agentorch 2.0.1` — OpenCode cloud client config and `AGENTS.md` updates are now built into the CLI. Install or upgrade with `pip install -U agentorch`.
+> **Latest release:** `agentorch 2.0.2` — MCP bridge with automatic local fallback, plus upgrade/reinstall docs for all IDE integrations. Install or upgrade with `pip install -U agentorch`.
 
 ```bash
-pip install -U agentorch>=2.0.0
+pip install -U agentorch>=2.0.2
 devtorch init
 devtorch setup          # connects to Claude Code, Cursor, or Antigravity
 devtorch debug timeline
@@ -78,13 +78,13 @@ devtorch debug timeline
 
 ### Reasoning Plus
 
-Reasoning Plus is DevTorch’s default augmentation mode: before every LLM call, it gathers smart context from your workspace and asks the model to produce inline `<thinking>` reasoning before the final answer. The reasoning is captured to `.GCC/` automatically and stripped from the response your tool sees.
+Reasoning Plus is DevTorch’s default augmentation mode: before every LLM call, it gathers smart context from your workspace, asks the model to produce inline `<thinking>` reasoning, and learns from previous calls by re-injecting relevant reasoning into subsequent prompts. The reasoning is captured to `.GCC/` automatically and stripped from the response your tool sees.
 
 ```bash
 devtorch reasoning-plus status          # show current settings
 devtorch reasoning-plus enable          # turn on per project
 devtorch reasoning-plus disable         # turn off per project
-devtorch reasoning-plus config --smart-top-n 3 --smart-max-lines 150
+devtorch reasoning-plus config --smart-top-n 3 --smart-max-lines 150 --learning-top-n 3
 ```
 
 Set `DEVTORCH_REASONING_PLUS=0` to disable it for a single session. See [REASONING_PLUS.md](REASONING_PLUS.md) for details.
@@ -113,6 +113,35 @@ pip install agentorch[all]       # everything
 
 ---
 
+## Upgrade from an older version
+
+DevTorch preserves your `.GCC/` store across upgrades. Newer releases add IDE configs, hooks, or governance instructions, so you should re-run `devtorch setup` after upgrading.
+
+**Upgrade any time:**
+
+```bash
+pip install -U agentorch>=2.0.2
+cd /path/to/your/project
+devtorch setup              # or: devtorch setup --target <ide>
+devtorch doctor
+```
+
+`devtorch setup` is idempotent:
+- It rewrites `CLAUDE.md` / `AGENTS.md` with the latest governance instructions.
+- It regenerates IDE-specific config files (`.claude/settings.json`, `.cursor/mcp.json`, `opencode.json`, etc.).
+- It leaves existing `.GCC/` data untouched.
+
+**If you use the cloud MCP server**, re-run the client config installer after upgrade to get the latest bridge and fallback settings:
+
+```bash
+export DEVTORCH_MCP_DOMAIN=devtorch-mcp.your-domain.com
+devtorch cloud client-config <ide> --org <org> --repo <repo> --api-key <api-key>
+```
+
+Supported `<ide>` values: `opencode`, `claude`, `cursor`, `vscode`, `stdio`.
+
+---
+
 ## Getting Started
 
 DevTorch takes **5 minutes to set up**. Pick your AI coding tool below.
@@ -155,6 +184,8 @@ Open Claude Code in your project. On the first task Claude will call `devtorch_c
 
 **What you get:** Full RACP integration — reasoning commits, sensitivity events, coordination vector Θ, divergence detection, HITL conflict resolution.
 
+> **Updating from an older version:** `pip install -U agentorch>=2.0.2 && devtorch setup` will rewrite `.claude/hooks.json`, `.claude/settings.json`, and `CLAUDE.md` with the latest versions.
+
 ---
 
 ### Cursor
@@ -197,6 +228,8 @@ devtorch doctor
 
 **What you get:** MCP tools, VS Code sidebar, optional HTTP proxy for passive capture of all LLM calls (`devtorch proxy start`).
 
+> **Updating from an older version:** `pip install -U agentorch>=2.0.2 && devtorch setup --target cursor` will regenerate `.cursor/mcp.json` and update `CLAUDE.md`.
+
 ---
 
 ### Google Antigravity
@@ -227,11 +260,13 @@ devtorch doctor
 
 **What you get:** Full RACP integration. The multi-agent nature of Antigravity makes Θ accumulation especially valuable — multiple agents share the same reasoning state through `devtorch_context`.
 
+> **Updating from an older version:** `pip install -U agentorch>=2.0.2 && devtorch setup --target antigravity` will regenerate the MCP config and update `AGENTS.md`.
+
 ---
 
 ### OpenCode
 
-OpenCode connects to the DevTorch cloud MCP server over SSE. It uses `opencode.json` for the MCP config and `AGENTS.md` for persistent governance instructions.
+OpenCode connects to DevTorch through the local stdio ↔ SSE bridge. The bridge talks to the cloud MCP server first, and if the cloud is unreachable or slow it automatically falls back to the local `devtorch` CLI so your `.GCC/` audit trail is never lost.
 
 **Step 1 — Install DevTorch**
 ```bash
@@ -251,7 +286,7 @@ devtorch cloud client-config opencode --org myorg --repo myrepo --api-key <api-k
 ```
 
 This creates:
-- `opencode.json` — the OpenCode MCP server config
+- `opencode.json` — the OpenCode MCP server config (uses the local bridge)
 - `AGENTS.md` — DevTorch governance instructions for the model
 
 **Step 3 — Verify**
@@ -261,7 +296,9 @@ devtorch doctor
 
 Open the project in OpenCode. The DevTorch MCP tools will be listed, and the model will be instructed to call `devtorch_context` at the start of each task and `devtorch_commit` before every file edit.
 
-**What you get:** Cloud-hosted MCP server integration, automatic reasoning commits via `AGENTS.md` instructions, and a shared Θ state across sessions.
+**What you get:** Cloud-hosted MCP server integration with automatic local fallback, automatic reasoning commits via `AGENTS.md` instructions, and a shared Θ state across sessions.
+
+> **Updating from an older version:** `pip install -U agentorch>=2.0.2 && devtorch cloud client-config opencode --org <org> --repo <repo> --api-key <api-key>` will regenerate `opencode.json` with the latest bridge and fallback settings. **Restart OpenCode** after updating the config.
 
 ---
 
@@ -293,6 +330,8 @@ devtorch sensitivity add --source you --concept auth --confidence 0.9 --disclosu
 
 **What you get:** Sidebar panel (read-only), git-commit-level event capture, manual reasoning commits. Full capture requires switching to Claude Code, Cursor, Kiro, or Antigravity.
 
+> **Updating from an older version:** `pip install -U agentorch>=2.0.2 && devtorch setup` will reinstall the git hook and update the governance instructions.
+
 ---
 
 ### AWS Kiro
@@ -356,6 +395,8 @@ devtorch doctor
 
 **What you get:** Full RACP integration equivalent to Claude Code — MCP tools, steering-document governance, agent action hooks.
 
+> **Updating from an older version:** `pip install -U agentorch>=2.0.2`, then re-run the manual steps above and refresh your steering document with the latest `AGENTS.md` text.
+
 ---
 
 ### AWS Bedrock (Python SDK)
@@ -387,6 +428,8 @@ All `invoke_model()` and `converse()` calls are identical. DevTorch injects the
 devtorch doctor
 ```
 
+> **Updating from an older version:** `pip install -U agentorch[wrapper]>=2.0.2`.
+
 ---
 
 ### Azure AI Foundry (Python SDK)
@@ -419,6 +462,8 @@ client = AzureOpenAI(
 devtorch doctor
 ```
 
+> **Updating from an older version:** `pip install -U agentorch[wrapper]>=2.0.2`.
+
 ---
 
 ### OpenAI Codex CLI
@@ -443,6 +488,8 @@ codex "Add error handling to the payment service"
 
 All Codex calls are intercepted by the proxy, which injects the RACP prefix and captures responses into `.GCC/`.
 
+> **Updating from an older version:** `pip install -U agentorch>=2.0.2`.
+
 ---
 
 ### Ollama (local models)
@@ -475,6 +522,8 @@ devtorch doctor
 
 No API key required. Fully local — RACP context and captures stay on-device.
 
+> **Updating from an older version:** `pip install -U agentorch[wrapper]>=2.0.2`.
+
 ---
 
 ### Cloud MCP server (team-wide shared governance)
@@ -516,6 +565,14 @@ Supported `<ide>` values: `opencode`, `claude`, `cursor`, `vscode`, `stdio`.
 
 **Full walkthrough:** `IDE_SETUP.md` has step-by-step setup for OpenCode, Claude Code, Cursor, VS Code, Kiro, Antigravity, JetBrains, and generic MCP clients.
 
+> **Updating from an older version:** You do **not** need to redeploy the Cloudflare Worker for client-side changes. Just regenerate the IDE config on each machine after upgrading `agentorch`:
+> ```bash
+> pip install -U agentorch>=2.0.2
+> export DEVTORCH_MCP_DOMAIN=devtorch-mcp.YOUR_ACCOUNT.workers.dev
+> /path/to/DevTorch/Implementation/deploy/cloudflare-worker/install-client-config.sh \
+>   <ide> <org> <repo> <api-key>
+> ```
+
 ---
 
 ### Platform comparison

{agentorch-2.0.1 → agentorch-2.0.2}/README.md

@@ -2,10 +2,10 @@
 
 **Local-first AI agent governance.** DevTorch gives multi-agent systems a shared on-disk audit trail, coordination vector, privacy controls, and LLM integration layer — all without a remote service.
 
-> **Latest release:** `agentorch 2.0.1` — OpenCode cloud client config and `AGENTS.md` updates are now built into the CLI. Install or upgrade with `pip install -U agentorch`.
+> **Latest release:** `agentorch 2.0.2` — MCP bridge with automatic local fallback, plus upgrade/reinstall docs for all IDE integrations. Install or upgrade with `pip install -U agentorch`.
 
 ```bash
-pip install -U agentorch>=2.0.0
+pip install -U agentorch>=2.0.2
 devtorch init
 devtorch setup          # connects to Claude Code, Cursor, or Antigravity
 devtorch debug timeline
@@ -28,13 +28,13 @@ devtorch debug timeline
 
 ### Reasoning Plus
 
-Reasoning Plus is DevTorch’s default augmentation mode: before every LLM call, it gathers smart context from your workspace and asks the model to produce inline `<thinking>` reasoning before the final answer. The reasoning is captured to `.GCC/` automatically and stripped from the response your tool sees.
+Reasoning Plus is DevTorch’s default augmentation mode: before every LLM call, it gathers smart context from your workspace, asks the model to produce inline `<thinking>` reasoning, and learns from previous calls by re-injecting relevant reasoning into subsequent prompts. The reasoning is captured to `.GCC/` automatically and stripped from the response your tool sees.
 
 ```bash
 devtorch reasoning-plus status          # show current settings
 devtorch reasoning-plus enable          # turn on per project
 devtorch reasoning-plus disable         # turn off per project
-devtorch reasoning-plus config --smart-top-n 3 --smart-max-lines 150
+devtorch reasoning-plus config --smart-top-n 3 --smart-max-lines 150 --learning-top-n 3
 ```
 
 Set `DEVTORCH_REASONING_PLUS=0` to disable it for a single session. See [REASONING_PLUS.md](REASONING_PLUS.md) for details.
@@ -63,6 +63,35 @@ pip install agentorch[all]       # everything
 
 ---
 
+## Upgrade from an older version
+
+DevTorch preserves your `.GCC/` store across upgrades. Newer releases add IDE configs, hooks, or governance instructions, so you should re-run `devtorch setup` after upgrading.
+
+**Upgrade any time:**
+
+```bash
+pip install -U agentorch>=2.0.2
+cd /path/to/your/project
+devtorch setup              # or: devtorch setup --target <ide>
+devtorch doctor
+```
+
+`devtorch setup` is idempotent:
+- It rewrites `CLAUDE.md` / `AGENTS.md` with the latest governance instructions.
+- It regenerates IDE-specific config files (`.claude/settings.json`, `.cursor/mcp.json`, `opencode.json`, etc.).
+- It leaves existing `.GCC/` data untouched.
+
+**If you use the cloud MCP server**, re-run the client config installer after upgrade to get the latest bridge and fallback settings:
+
+```bash
+export DEVTORCH_MCP_DOMAIN=devtorch-mcp.your-domain.com
+devtorch cloud client-config <ide> --org <org> --repo <repo> --api-key <api-key>
+```
+
+Supported `<ide>` values: `opencode`, `claude`, `cursor`, `vscode`, `stdio`.
+
+---
+
 ## Getting Started
 
 DevTorch takes **5 minutes to set up**. Pick your AI coding tool below.
@@ -105,6 +134,8 @@ Open Claude Code in your project. On the first task Claude will call `devtorch_c
 
 **What you get:** Full RACP integration — reasoning commits, sensitivity events, coordination vector Θ, divergence detection, HITL conflict resolution.
 
+> **Updating from an older version:** `pip install -U agentorch>=2.0.2 && devtorch setup` will rewrite `.claude/hooks.json`, `.claude/settings.json`, and `CLAUDE.md` with the latest versions.
+
 ---
 
 ### Cursor
@@ -147,6 +178,8 @@ devtorch doctor
 
 **What you get:** MCP tools, VS Code sidebar, optional HTTP proxy for passive capture of all LLM calls (`devtorch proxy start`).
 
+> **Updating from an older version:** `pip install -U agentorch>=2.0.2 && devtorch setup --target cursor` will regenerate `.cursor/mcp.json` and update `CLAUDE.md`.
+
 ---
 
 ### Google Antigravity
@@ -177,11 +210,13 @@ devtorch doctor
 
 **What you get:** Full RACP integration. The multi-agent nature of Antigravity makes Θ accumulation especially valuable — multiple agents share the same reasoning state through `devtorch_context`.
 
+> **Updating from an older version:** `pip install -U agentorch>=2.0.2 && devtorch setup --target antigravity` will regenerate the MCP config and update `AGENTS.md`.
+
 ---
 
 ### OpenCode
 
-OpenCode connects to the DevTorch cloud MCP server over SSE. It uses `opencode.json` for the MCP config and `AGENTS.md` for persistent governance instructions.
+OpenCode connects to DevTorch through the local stdio ↔ SSE bridge. The bridge talks to the cloud MCP server first, and if the cloud is unreachable or slow it automatically falls back to the local `devtorch` CLI so your `.GCC/` audit trail is never lost.
 
 **Step 1 — Install DevTorch**
 ```bash
@@ -201,7 +236,7 @@ devtorch cloud client-config opencode --org myorg --repo myrepo --api-key <api-k
 ```
 
 This creates:
-- `opencode.json` — the OpenCode MCP server config
+- `opencode.json` — the OpenCode MCP server config (uses the local bridge)
 - `AGENTS.md` — DevTorch governance instructions for the model
 
 **Step 3 — Verify**
@@ -211,7 +246,9 @@ devtorch doctor
 
 Open the project in OpenCode. The DevTorch MCP tools will be listed, and the model will be instructed to call `devtorch_context` at the start of each task and `devtorch_commit` before every file edit.
 
-**What you get:** Cloud-hosted MCP server integration, automatic reasoning commits via `AGENTS.md` instructions, and a shared Θ state across sessions.
+**What you get:** Cloud-hosted MCP server integration with automatic local fallback, automatic reasoning commits via `AGENTS.md` instructions, and a shared Θ state across sessions.
+
+> **Updating from an older version:** `pip install -U agentorch>=2.0.2 && devtorch cloud client-config opencode --org <org> --repo <repo> --api-key <api-key>` will regenerate `opencode.json` with the latest bridge and fallback settings. **Restart OpenCode** after updating the config.
 
 ---
 
@@ -243,6 +280,8 @@ devtorch sensitivity add --source you --concept auth --confidence 0.9 --disclosu
 
 **What you get:** Sidebar panel (read-only), git-commit-level event capture, manual reasoning commits. Full capture requires switching to Claude Code, Cursor, Kiro, or Antigravity.
 
+> **Updating from an older version:** `pip install -U agentorch>=2.0.2 && devtorch setup` will reinstall the git hook and update the governance instructions.
+
 ---
 
 ### AWS Kiro
@@ -306,6 +345,8 @@ devtorch doctor
 
 **What you get:** Full RACP integration equivalent to Claude Code — MCP tools, steering-document governance, agent action hooks.
 
+> **Updating from an older version:** `pip install -U agentorch>=2.0.2`, then re-run the manual steps above and refresh your steering document with the latest `AGENTS.md` text.
+
 ---
 
 ### AWS Bedrock (Python SDK)
@@ -337,6 +378,8 @@ All `invoke_model()` and `converse()` calls are identical. DevTorch injects the
 devtorch doctor
 ```
 
+> **Updating from an older version:** `pip install -U agentorch[wrapper]>=2.0.2`.
+
 ---
 
 ### Azure AI Foundry (Python SDK)
@@ -369,6 +412,8 @@ client = AzureOpenAI(
 devtorch doctor
 ```
 
+> **Updating from an older version:** `pip install -U agentorch[wrapper]>=2.0.2`.
+
 ---
 
 ### OpenAI Codex CLI
@@ -393,6 +438,8 @@ codex "Add error handling to the payment service"
 
 All Codex calls are intercepted by the proxy, which injects the RACP prefix and captures responses into `.GCC/`.
 
+> **Updating from an older version:** `pip install -U agentorch>=2.0.2`.
+
 ---
 
 ### Ollama (local models)
@@ -425,6 +472,8 @@ devtorch doctor
 
 No API key required. Fully local — RACP context and captures stay on-device.
 
+> **Updating from an older version:** `pip install -U agentorch[wrapper]>=2.0.2`.
+
 ---
 
 ### Cloud MCP server (team-wide shared governance)
@@ -466,6 +515,14 @@ Supported `<ide>` values: `opencode`, `claude`, `cursor`, `vscode`, `stdio`.
 
 **Full walkthrough:** `IDE_SETUP.md` has step-by-step setup for OpenCode, Claude Code, Cursor, VS Code, Kiro, Antigravity, JetBrains, and generic MCP clients.
 
+> **Updating from an older version:** You do **not** need to redeploy the Cloudflare Worker for client-side changes. Just regenerate the IDE config on each machine after upgrading `agentorch`:
+> ```bash
+> pip install -U agentorch>=2.0.2
+> export DEVTORCH_MCP_DOMAIN=devtorch-mcp.YOUR_ACCOUNT.workers.dev
+> /path/to/DevTorch/Implementation/deploy/cloudflare-worker/install-client-config.sh \
+>   <ide> <org> <repo> <api-key>
+> ```
+
 ---
 
 ### Platform comparison

{agentorch-2.0.1 → agentorch-2.0.2}/agentorch.egg-info/PKG-INFO

@@ -1,6 +1,6 @@
 Metadata-Version: 2.4
 Name: agentorch
-Version: 2.0.1
+Version: 2.0.2
 Summary: DevTorch — AI reasoning capture, audit trail, and governance for agent-assisted development
 Author-email: Hemant Joshi <hemant@flotorch.ai>
 License: Apache-2.0
@@ -52,10 +52,10 @@ Requires-Dist: agentorch[cloud,mcp,proxy,wrapper]; extra == "all"
 
 **Local-first AI agent governance.** DevTorch gives multi-agent systems a shared on-disk audit trail, coordination vector, privacy controls, and LLM integration layer — all without a remote service.
 
-> **Latest release:** `agentorch 2.0.1` — OpenCode cloud client config and `AGENTS.md` updates are now built into the CLI. Install or upgrade with `pip install -U agentorch`.
+> **Latest release:** `agentorch 2.0.2` — MCP bridge with automatic local fallback, plus upgrade/reinstall docs for all IDE integrations. Install or upgrade with `pip install -U agentorch`.
 
 ```bash
-pip install -U agentorch>=2.0.0
+pip install -U agentorch>=2.0.2
 devtorch init
 devtorch setup          # connects to Claude Code, Cursor, or Antigravity
 devtorch debug timeline
@@ -78,13 +78,13 @@ devtorch debug timeline
 
 ### Reasoning Plus
 
-Reasoning Plus is DevTorch’s default augmentation mode: before every LLM call, it gathers smart context from your workspace and asks the model to produce inline `<thinking>` reasoning before the final answer. The reasoning is captured to `.GCC/` automatically and stripped from the response your tool sees.
+Reasoning Plus is DevTorch’s default augmentation mode: before every LLM call, it gathers smart context from your workspace, asks the model to produce inline `<thinking>` reasoning, and learns from previous calls by re-injecting relevant reasoning into subsequent prompts. The reasoning is captured to `.GCC/` automatically and stripped from the response your tool sees.
 
 ```bash
 devtorch reasoning-plus status          # show current settings
 devtorch reasoning-plus enable          # turn on per project
 devtorch reasoning-plus disable         # turn off per project
-devtorch reasoning-plus config --smart-top-n 3 --smart-max-lines 150
+devtorch reasoning-plus config --smart-top-n 3 --smart-max-lines 150 --learning-top-n 3
 ```
 
 Set `DEVTORCH_REASONING_PLUS=0` to disable it for a single session. See [REASONING_PLUS.md](REASONING_PLUS.md) for details.
@@ -113,6 +113,35 @@ pip install agentorch[all]       # everything
 
 ---
 
+## Upgrade from an older version
+
+DevTorch preserves your `.GCC/` store across upgrades. Newer releases add IDE configs, hooks, or governance instructions, so you should re-run `devtorch setup` after upgrading.
+
+**Upgrade any time:**
+
+```bash
+pip install -U agentorch>=2.0.2
+cd /path/to/your/project
+devtorch setup              # or: devtorch setup --target <ide>
+devtorch doctor
+```
+
+`devtorch setup` is idempotent:
+- It rewrites `CLAUDE.md` / `AGENTS.md` with the latest governance instructions.
+- It regenerates IDE-specific config files (`.claude/settings.json`, `.cursor/mcp.json`, `opencode.json`, etc.).
+- It leaves existing `.GCC/` data untouched.
+
+**If you use the cloud MCP server**, re-run the client config installer after upgrade to get the latest bridge and fallback settings:
+
+```bash
+export DEVTORCH_MCP_DOMAIN=devtorch-mcp.your-domain.com
+devtorch cloud client-config <ide> --org <org> --repo <repo> --api-key <api-key>
+```
+
+Supported `<ide>` values: `opencode`, `claude`, `cursor`, `vscode`, `stdio`.
+
+---
+
 ## Getting Started
 
 DevTorch takes **5 minutes to set up**. Pick your AI coding tool below.
@@ -155,6 +184,8 @@ Open Claude Code in your project. On the first task Claude will call `devtorch_c
 
 **What you get:** Full RACP integration — reasoning commits, sensitivity events, coordination vector Θ, divergence detection, HITL conflict resolution.
 
+> **Updating from an older version:** `pip install -U agentorch>=2.0.2 && devtorch setup` will rewrite `.claude/hooks.json`, `.claude/settings.json`, and `CLAUDE.md` with the latest versions.
+
 ---
 
 ### Cursor
@@ -197,6 +228,8 @@ devtorch doctor
 
 **What you get:** MCP tools, VS Code sidebar, optional HTTP proxy for passive capture of all LLM calls (`devtorch proxy start`).
 
+> **Updating from an older version:** `pip install -U agentorch>=2.0.2 && devtorch setup --target cursor` will regenerate `.cursor/mcp.json` and update `CLAUDE.md`.
+
 ---
 
 ### Google Antigravity
@@ -227,11 +260,13 @@ devtorch doctor
 
 **What you get:** Full RACP integration. The multi-agent nature of Antigravity makes Θ accumulation especially valuable — multiple agents share the same reasoning state through `devtorch_context`.
 
+> **Updating from an older version:** `pip install -U agentorch>=2.0.2 && devtorch setup --target antigravity` will regenerate the MCP config and update `AGENTS.md`.
+
 ---
 
 ### OpenCode
 
-OpenCode connects to the DevTorch cloud MCP server over SSE. It uses `opencode.json` for the MCP config and `AGENTS.md` for persistent governance instructions.
+OpenCode connects to DevTorch through the local stdio ↔ SSE bridge. The bridge talks to the cloud MCP server first, and if the cloud is unreachable or slow it automatically falls back to the local `devtorch` CLI so your `.GCC/` audit trail is never lost.
 
 **Step 1 — Install DevTorch**
 ```bash
@@ -251,7 +286,7 @@ devtorch cloud client-config opencode --org myorg --repo myrepo --api-key <api-k
 ```
 
 This creates:
-- `opencode.json` — the OpenCode MCP server config
+- `opencode.json` — the OpenCode MCP server config (uses the local bridge)
 - `AGENTS.md` — DevTorch governance instructions for the model
 
 **Step 3 — Verify**
@@ -261,7 +296,9 @@ devtorch doctor
 
 Open the project in OpenCode. The DevTorch MCP tools will be listed, and the model will be instructed to call `devtorch_context` at the start of each task and `devtorch_commit` before every file edit.
 
-**What you get:** Cloud-hosted MCP server integration, automatic reasoning commits via `AGENTS.md` instructions, and a shared Θ state across sessions.
+**What you get:** Cloud-hosted MCP server integration with automatic local fallback, automatic reasoning commits via `AGENTS.md` instructions, and a shared Θ state across sessions.
+
+> **Updating from an older version:** `pip install -U agentorch>=2.0.2 && devtorch cloud client-config opencode --org <org> --repo <repo> --api-key <api-key>` will regenerate `opencode.json` with the latest bridge and fallback settings. **Restart OpenCode** after updating the config.
 
 ---
 
@@ -293,6 +330,8 @@ devtorch sensitivity add --source you --concept auth --confidence 0.9 --disclosu
 
 **What you get:** Sidebar panel (read-only), git-commit-level event capture, manual reasoning commits. Full capture requires switching to Claude Code, Cursor, Kiro, or Antigravity.
 
+> **Updating from an older version:** `pip install -U agentorch>=2.0.2 && devtorch setup` will reinstall the git hook and update the governance instructions.
+
 ---
 
 ### AWS Kiro
@@ -356,6 +395,8 @@ devtorch doctor
 
 **What you get:** Full RACP integration equivalent to Claude Code — MCP tools, steering-document governance, agent action hooks.
 
+> **Updating from an older version:** `pip install -U agentorch>=2.0.2`, then re-run the manual steps above and refresh your steering document with the latest `AGENTS.md` text.
+
 ---
 
 ### AWS Bedrock (Python SDK)
@@ -387,6 +428,8 @@ All `invoke_model()` and `converse()` calls are identical. DevTorch injects the
 devtorch doctor
 ```
 
+> **Updating from an older version:** `pip install -U agentorch[wrapper]>=2.0.2`.
+
 ---
 
 ### Azure AI Foundry (Python SDK)
@@ -419,6 +462,8 @@ client = AzureOpenAI(
 devtorch doctor
 ```
 
+> **Updating from an older version:** `pip install -U agentorch[wrapper]>=2.0.2`.
+
 ---
 
 ### OpenAI Codex CLI
@@ -443,6 +488,8 @@ codex "Add error handling to the payment service"
 
 All Codex calls are intercepted by the proxy, which injects the RACP prefix and captures responses into `.GCC/`.
 
+> **Updating from an older version:** `pip install -U agentorch>=2.0.2`.
+
 ---
 
 ### Ollama (local models)
@@ -475,6 +522,8 @@ devtorch doctor
 
 No API key required. Fully local — RACP context and captures stay on-device.
 
+> **Updating from an older version:** `pip install -U agentorch[wrapper]>=2.0.2`.
+
 ---
 
 ### Cloud MCP server (team-wide shared governance)
@@ -516,6 +565,14 @@ Supported `<ide>` values: `opencode`, `claude`, `cursor`, `vscode`, `stdio`.
 
 **Full walkthrough:** `IDE_SETUP.md` has step-by-step setup for OpenCode, Claude Code, Cursor, VS Code, Kiro, Antigravity, JetBrains, and generic MCP clients.
 
+> **Updating from an older version:** You do **not** need to redeploy the Cloudflare Worker for client-side changes. Just regenerate the IDE config on each machine after upgrading `agentorch`:
+> ```bash
+> pip install -U agentorch>=2.0.2
+> export DEVTORCH_MCP_DOMAIN=devtorch-mcp.YOUR_ACCOUNT.workers.dev
+> /path/to/DevTorch/Implementation/deploy/cloudflare-worker/install-client-config.sh \
+>   <ide> <org> <repo> <api-key>
+> ```
+
 ---
 
 ### Platform comparison

{agentorch-2.0.1 → agentorch-2.0.2}/agentorch.egg-info/SOURCES.txt

@@ -18,6 +18,7 @@ benchmarks/swe_bench/run_baseline.py
 benchmarks/swe_bench/run_devtorch.py
 benchmarks/swe_bench/seed_gcc.py
 devtorch_cli/__init__.py
+devtorch_cli/__main__.py
 devtorch_cli/main.py
 devtorch_core/__init__.py
 devtorch_core/aggphi_textual.py
@@ -131,6 +132,15 @@ devtorch_core/reasoning_plus/capture.py
 devtorch_core/reasoning_plus/config.py
 devtorch_core/reasoning_plus/context.py
 devtorch_core/reasoning_plus/prompt.py
+devtorch_core/reasoning_plus/learning/__init__.py
+devtorch_core/reasoning_plus/learning/api.py
+devtorch_core/reasoning_plus/learning/composer.py
+devtorch_core/reasoning_plus/learning/extractor.py
+devtorch_core/reasoning_plus/learning/models.py
+devtorch_core/reasoning_plus/learning/recorder.py
+devtorch_core/reasoning_plus/learning/relevance.py
+devtorch_core/reasoning_plus/learning/state.py
+devtorch_core/reasoning_plus/learning/store.py
 devtorch_core/rep_network/__init__.py
 devtorch_core/rep_network/merge.py
 devtorch_core/rep_network/node.py
@@ -153,6 +163,7 @@ tests/conftest.py
 tests/test_backend_factory.py
 tests/test_broadcast.py
 tests/test_cloud_auth.py
+tests/test_cloud_bridge.py
 tests/test_cloud_integration.py
 tests/test_cloud_sse.py
 tests/test_cloud_storage.py
@@ -190,6 +201,7 @@ tests/test_s21_hitl.py
 tests/test_s22_swe_bench.py
 tests/test_s22_swe_bench_dry_run.py
 tests/test_s23_reasoning_plus.py
+tests/test_s24_reasoning_plus_learning.py
 tests/test_s2_context.py
 tests/test_s3_sensitivity.py
 tests/test_s4_invariants.py

{agentorch-2.0.1 → agentorch-2.0.2}/benchmarks/swe_bench/patch_utils.py

@@ -9,6 +9,10 @@ import shutil
 import subprocess
 import tempfile
 from pathlib import Path
+from typing import TYPE_CHECKING
+
+if TYPE_CHECKING:
+    from devtorch_core.reasoning_plus.learning import ReasoningPlusLearning
 
 
 def extract_file_paths_from_patch(patch_text: str) -> list[str]:
@@ -292,15 +296,21 @@ def self_correct_patch(
     base_commit: str,
     original_patch: str,
     max_retries: int = 2,
+    learning: "ReasoningPlusLearning | None" = None,
+    session_id: str = "",
 ) -> tuple[str, int, bool, str]:
     """
     Try to fix a malformed patch by asking the model again.
 
+    If a `learning` object is provided, each failed attempt is recorded and the
+    distilled learnings are injected into the next correction prompt.
+
     Returns (final_patch, attempts_used, applies_cleanly, last_error).
     """
     patch = normalize_patch(original_patch)
     attempts = 0
     last_error = ""
+    previous_error = ""
 
     for attempt in range(1, max_retries + 1):
         temp_repo = None
@@ -311,6 +321,23 @@ def self_correct_patch(
                 return patch, attempts, True, ""
             last_error = error
 
+            # Record the failed patch application as a learning opportunity.
+            if learning and previous_error != error:
+                try:
+                    call_id = learning.record_call(
+                        call_type="tool",
+                        name="git_apply_check",
+                        reasoning=f"The patch failed to apply: {error}",
+                        input=patch,
+                        output=error,
+                        outcome="failure",
+                        session_id=session_id,
+                    )
+                    learning.extract_learnings(call_id)
+                except Exception:
+                    pass
+                previous_error = error
+
             correction_prompt = (
                 "Your previous patch failed to apply to the repository. "
                 "Please produce a corrected unified diff patch that applies cleanly. "
@@ -318,6 +345,18 @@ def self_correct_patch(
                 f"Apply error:\n{error}\n\n"
                 "Original issue and files are above."
             )
+
+            # Inject relevant learnings from previous attempts into the correction prompt.
+            if learning:
+                try:
+                    learnings = learning.get_relevant_learnings(
+                        correction_prompt,
+                        top_n=learning._config.learning_top_n,
+                    )
+                    correction_prompt = learning.compose_user_prompt(correction_prompt, learnings)
+                except Exception:
+                    pass
+
             response = client.messages.create(
                 model=model,
                 max_tokens=4096,