build: purge obsolete rca docs and redundant installer scripts
Some checks failed
Deploy-Agent-V15-Stdin / JOB-V15-STDIN (push) Failing after 3s
Some checks failed
Deploy-Agent-V15-Stdin / JOB-V15-STDIN (push) Failing after 3s
- Deleted docs/rca/ as they were early v0.1.0 development artifacts. - Deleted minimal.asd. - Deleted scripts/onboard-baremetal.sh (subsumed by opencortex.sh). - Moved scripts/browser-bridge.py to skills/assets/ for upcoming Web Research skill. - Removed scripts/ directory.
This commit is contained in:
@@ -1,42 +0,0 @@
|
||||
#+TITLE: Root Cause Analysis: Micro-Loader & Deterministic Boot Sequence
|
||||
#+DATE: 2026-04-11
|
||||
#+FILETAGS: :rca:boot:loader:topological-sort:autonomy:
|
||||
|
||||
* Executive Summary
|
||||
Refactored the arbitrary skill loading mechanism into a robust **Micro-Loader**. The system now calculates a deterministic boot sequence based on `#+DEPENDS_ON:` tags and protects the harness from malformed or hanging skills via package-based jailing and execution timeouts.
|
||||
|
||||
* 1. Issue: Fragile Load Order
|
||||
** Symptoms
|
||||
Skills that depended on functions or variables from other skills would randomly fail to load depending on the filesystem's directory traversal order.
|
||||
** Root Cause
|
||||
`initialize-all-skills` used a simple `dolist` over `uiop:directory-files`, which has no semantic awareness of inter-skill dependencies.
|
||||
** Resolution
|
||||
1. **Metadata Scanning:** Implemented `parse-skill-metadata` to extract `:ID:` and `#+DEPENDS_ON:` without executing code.
|
||||
2. **Topological Sort:** Implemented a DFS-based `topological-sort-skills` to guarantee that prerequisites are loaded before their dependents.
|
||||
3. **Circular Detection:** Added explicit detection and error reporting for circular dependency loops.
|
||||
|
||||
* 2. Issue: Shared State Corruption (Brain Rot)
|
||||
** Symptoms
|
||||
Variables or functions with the same name in different skills would silently overwrite each other, causing unpredictable behavior.
|
||||
** Root Cause
|
||||
All skills were being evaluated directly into the `opencortex` package.
|
||||
** Resolution
|
||||
**Package-Based Jailing:** Each skill is now evaluated within its own dedicated, shadowed package (e.g., `OPENCORTEX.SKILLS.ORG-SKILL-CHAT`). This ensures logical isolation while still allowing access to kernel exports.
|
||||
|
||||
* 3. Issue: Boot Stall (The Hanging Skill)
|
||||
** Symptoms
|
||||
A single skill with an infinite loop or heavy synchronous initialization could hang the entire agent during startup.
|
||||
** Root Cause
|
||||
Skill loading was strictly synchronous and blocking on the main thread.
|
||||
** Resolution
|
||||
**Execution Timeouts:** Implemented `load-skill-with-timeout`, which wraps the loader in a monitored thread. If a skill takes longer than 5 seconds to initialize, the loader terminates the thread, jails the failure, and continues with the rest of the boot sequence.
|
||||
|
||||
* 4. opencortex Mandate Alignment
|
||||
** Evolutionary Kernel
|
||||
The boot sequence is now a verifiable, mathematical process rather than a side-effect of filesystem organization.
|
||||
** Literate Granularity
|
||||
The `org-skill-skills.org` source was refactored into a strictly granular "one definition per block" format.
|
||||
|
||||
* 5. Permanent Learnings
|
||||
- **Reverse Topological Order:** Remember that a DFS-based sort with `push` needs an `nreverse` to place dependencies at the front of the list.
|
||||
- **Path Portability:** Use `uiop:getcwd` instead of `pwd` for more reliable path resolution across different Lisp implementations and OSes.
|
||||
@@ -1,33 +0,0 @@
|
||||
#+TITLE: Root Cause Analysis: Deterministic Engine Bouncer & Authorization Gate
|
||||
#+DATE: 2026-04-11
|
||||
#+FILETAGS: :rca:bouncer:authorization:autonomy:security:
|
||||
|
||||
* Executive Summary
|
||||
Implemented the "Planning Mode" Bouncer to intercept high-risk Probabilistic Engine proposals (e.g., shell commands, Lisp evaluation). The system now forces these actions into an asynchronous "Flight Plan" Org node for manual Autonomous approval, fulfilling the "everything is a node" and high-integrity mandates.
|
||||
|
||||
* 1. Issue: Automated High-Risk Execution
|
||||
** Symptoms
|
||||
Probabilistic Engine proposals involving `shell` or `eval` were executed immediately upon passing the `decide` gate's safety harness. This lacked human-in-the-loop oversight for irreversible or complex operations.
|
||||
** Root Cause
|
||||
Architecture gap. The system lacked an authorization state between "Safe" and "Executed".
|
||||
** Resolution
|
||||
1. **Interceptor:** Added `bouncer-check` to `deterministic.lisp`. It flags high-risk actions that lack the `:approved t` property.
|
||||
2. **Asynchronous Event:** If flagged, the harness emits an `:approval-required` event.
|
||||
3. **Flight Plan Skill:** Created `org-skill-bouncer.org` to:
|
||||
- Catch the event and create a serialized Org node with state `PLAN`.
|
||||
- Monitor the Memory for `APPROVED` states.
|
||||
- Re-inject approved actions with the `:approved t` bypass flag.
|
||||
|
||||
* 2. Design Decision: Org-native Approval
|
||||
** Requirement
|
||||
Align with "Homoiconic Memory" and "Lisp Machine Autonomousty".
|
||||
** Selected Path
|
||||
State-Based Approval (Org-native).
|
||||
- *Pros:* Auditable, asynchronous, utilizes existing Org-mode workflows.
|
||||
- *Cons:* Slightly more latency than an interactive prompt.
|
||||
** Alignment
|
||||
Ensures that the agent's "Flight Plans" are first-class citizens in the Memex, allowing the Autonomous to review and approve them using standard GTD tools.
|
||||
|
||||
* 3. Permanent Learnings
|
||||
- **Serial Bypass:** Always include a specific bypass flag (e.g., `:approved t`) when re-injecting intercepted actions to prevent infinite interception loops.
|
||||
- **Heartbeat Listeners:** Periodic scanning of the Memory for state transitions is an effective way to implement asynchronous authorization gates without blocking the harness.
|
||||
@@ -1,36 +0,0 @@
|
||||
#+TITLE: Root Cause Analysis: Lisp-Native Formal Verification Gate
|
||||
#+DATE: 2026-04-11
|
||||
#+FILETAGS: :rca:security:formal-verification:autonomy:
|
||||
|
||||
* Executive Summary
|
||||
Implemented a Lisp-Native Deterministic Prover to replace heuristic whitelisting with formal security invariants. This ensures that every high-impact action (shell, file I/O) is mathematically proven safe against the Autonomous's core mandates.
|
||||
|
||||
* 1. Architectural Shift: Native vs. External
|
||||
** Issue
|
||||
The initial draft suggested using `Z3`, an external SMT solver. However, `Z3` was not available in the environment and would add significant complexity/bloat to the Docker image.
|
||||
** Resolution
|
||||
Leveraged Common Lisp's inherent strength in symbol manipulation to build a **Lisp-Native Prover**. Invariants are defined as high-order predicates that operate on the structure of proposed actions. This provides a self-contained, high-performance verification layer.
|
||||
|
||||
* 2. Issue: Dependency Fragility
|
||||
** Symptoms
|
||||
System failed to load with `Package STR does not exist`.
|
||||
** Root Cause
|
||||
Incorrect assumption about the Quicklisp system name vs. the package name. The library is `cl-str` but the Quicklisp system is `str` and the package is `str`.
|
||||
** Resolution
|
||||
1. Updated `opencortex.asd` to depend on `:str`.
|
||||
2. Updated all source code and literate notes to use the `str:` prefix.
|
||||
3. Verified via explicit `ql:quickload` in the test runner.
|
||||
|
||||
* 3. Formal Invariants Implemented
|
||||
- **Path Confinement:** Deterministically proves that any file operation or absolute path in a shell command is strictly within the `/home/user/memex` root.
|
||||
- **No Network Exfiltration:** Prevents the shell from invoking common exfiltration tools (`nc`, `ssh`, etc.) by inspecting the parsed command structure.
|
||||
|
||||
* 4. opencortex Mandate Alignment
|
||||
** Soundness over Heuristics
|
||||
By moving to formal invariants, we have moved from "blacklisting bad things" to "proving safety." Any action that cannot be proven to satisfy all invariants is denied by default.
|
||||
** Literate Granularity
|
||||
The `org-skill-formal-verification.org` file follows the "one definition per block" mandate, ensuring that the logic of each invariant is individually documented and verifiable.
|
||||
|
||||
* 5. Permanent Learnings
|
||||
- **Tooling Independence:** Whenever possible, prefer native Lisp logic over external binaries for core security gates to reduce the attack surface and deployment complexity.
|
||||
- **Environment Consistency:** Always use `(setf (uiop:getenv ...) ...)` for portable environment manipulation in tests.
|
||||
@@ -1,40 +0,0 @@
|
||||
#+TITLE: Root Cause Analysis: Matrix Gateway & Communication Track Completion
|
||||
#+DATE: 2026-04-11
|
||||
#+FILETAGS: :rca:gateway:matrix:chat:autonomy:
|
||||
|
||||
* Executive Summary
|
||||
Successfully implemented the third and final external communication channel (Matrix) for OpenCortex v1.0. Resolved integration issues related to case-sensitivity in JSON keys and strict header requirements in `dexador`.
|
||||
|
||||
* 1. Issue: Symbol Casing in JSON Keys
|
||||
** Symptoms
|
||||
The `TEST-MATRIX-INBOUND-NORMALIZATION` test failed because `room-id` was being extracted as `"!ROOM:HS.ORG"` (uppercase) instead of `"!room:hs.org"`.
|
||||
** Root Cause
|
||||
Common Lisp's default reader converts symbol names to uppercase. When `(string car-of-alist)` was called on a symbol generated by `cl-json`, it produced an uppercase string.
|
||||
** Resolution
|
||||
Updated the implementation to use `(string-downcase (string ...))` for room IDs and other case-sensitive Matrix identifiers.
|
||||
|
||||
* 2. Issue: Since Token Extraction Failure
|
||||
** Symptoms
|
||||
The sync loop failed to update the `*matrix-since-token*`, causing duplicate message processing risk.
|
||||
** Root Cause
|
||||
Anticipating `:next-batch` but receiving `:next--batch` (or vice versa) due to inconsistent `cl-json` behavior across different environments or structures.
|
||||
** Resolution
|
||||
Implemented a robust `(or (cdr (assoc :next-batch json)) (cdr (assoc :next--batch json)))` lookup to handle both hyphenation styles.
|
||||
|
||||
* 3. Issue: Type Error in Authorization Headers
|
||||
** Symptoms
|
||||
`dex:put` crashed with a `TYPE-ERROR`.
|
||||
** Root Cause
|
||||
I was passing a single string or an incorrectly nested list where `dexador` expected a strict alist of header pairs `(("Key" . "Value") ...)`.
|
||||
** Resolution
|
||||
Standardized all gateway HTTP calls to use proper alist nesting for headers.
|
||||
|
||||
* 4. Completion: Communication Track
|
||||
With Telegram, Signal, and Matrix gateways now verified and passing tests, the OpenCortex has achieved full multi-channel parity.
|
||||
- **Telegram:** Polling via Bot API.
|
||||
- **Signal:** Wrapping `signal-cli`.
|
||||
- **Matrix:** Polling via `/sync` Client API.
|
||||
|
||||
* 5. Permanent Learnings
|
||||
- **Case Sensitivity:** Matrix IDs (rooms, users) are case-sensitive; Lisp symbols are not. Always force downcasing or use strings for storage.
|
||||
- **Header Alists:** Always use dotted pairs `("Key" . "Value")` for `dexador` headers.
|
||||
@@ -1,33 +0,0 @@
|
||||
#+TITLE: Root Cause Analysis: Signal Gateway & Multi-Channel Chat
|
||||
#+DATE: 2026-04-11
|
||||
#+FILETAGS: :rca:gateway:signal:chat:autonomy:
|
||||
|
||||
* Executive Summary
|
||||
Successfully implemented the second external communication channel (Signal) using `signal-cli`. Further hardened the multi-channel chat logic and resolved JSON mapping discrepancies between Common Lisp and external CLI outputs.
|
||||
|
||||
* 1. Issue: JSON Key Mapping Mismatch
|
||||
** Symptoms
|
||||
The `TEST-SIGNAL-INBOUND-NORMALIZATION` test failed despite the mock JSON appearing correct.
|
||||
** Root Cause
|
||||
`cl-json` default behavior for decoding. It converts camelCase keys from JSON (e.g., `dataMessage`) into kebab-case keywords in Lisp (e.g., `:DATA-MESSAGE`). I had incorrectly anticipated `:DATA--MESSAGE` or `:DATA_MESSAGE`.
|
||||
** Resolution
|
||||
1. **Diagnostic:** Added debug output to the test suite to inspect the exact plist structure returned by `cl-json`.
|
||||
2. **Correction:** Updated both the implementation and the literate note to use the correct `:DATA-MESSAGE` and `:SOURCE` keywords.
|
||||
|
||||
* 2. Implementation: Signal-CLI Wrapper
|
||||
** Strategy
|
||||
Unlike Telegram's HTTP API, Signal requires a local binary (`signal-cli`).
|
||||
- **Sensor:** Uses `uiop:run-program` with `receive --json` in a polling loop (5s interval).
|
||||
- **Actuator:** Uses `uiop:run-program` with `send -m <text> <recipient>`.
|
||||
** Security
|
||||
The system uses the pre-configured Signal account `+13322690326` discovered in the user's memex.
|
||||
|
||||
* 3. Alignment with opencortex Mandates
|
||||
** Literate Granularity
|
||||
Strictly adhered to the "one definition per block" mandate throughout the new `org-skill-gateway-signal.org` file.
|
||||
** Verification
|
||||
The `gateway-signal-suite` (10 checks) provides full coverage for inbound parsing and outbound command generation.
|
||||
|
||||
* 4. Permanent Learnings
|
||||
- **JSON Semantics:** Always verify the specific keyword transformation rules of the JSON library when dealing with external CLI outputs.
|
||||
- **Process Robustness:** `uiop:run-program` is the reliable standard for CLI-based gateways in SBCL.
|
||||
@@ -1,43 +0,0 @@
|
||||
#+TITLE: Root Cause Analysis: Telegram Gateway & Channel-Aware Chat
|
||||
#+DATE: 2026-04-11
|
||||
#+FILETAGS: :rca:gateway:telegram:chat:autonomy:
|
||||
|
||||
* Executive Summary
|
||||
Successfully implemented the first external communication channel (Telegram) and decoupled the Chat Agent from its Emacs-centric roots. Resolved significant load-order and dependency issues identified during integration.
|
||||
|
||||
* 1. Issue: Undefined Foundational Functions
|
||||
** Symptoms
|
||||
During compilation, `gateway-telegram.lisp` failed with `UNDEFINED-FUNCTION` for `register-actuator` and `harness-log`.
|
||||
** Root Cause
|
||||
Poorly scoped foundational functions. These were defined in `core.lisp` (the loop orchestrator), which was loaded *after* the gateways in `opencortex.asd`. This created a "Circular Intention" where the gateways needed the harness to exist before the harness could load the gateways.
|
||||
** Resolution
|
||||
1. **Relocation:** Moved `*actuator-registry*` and `register-actuator` to `communication.lisp` (the foundation).
|
||||
2. **Reordering:** Adjusted `opencortex.asd` to load `core.lisp` (containing the stimulus loop) immediately after the deterministic gates but before the physical sensors (gateways).
|
||||
|
||||
* 2. Issue: Hardcoded Chat UI
|
||||
** Symptoms
|
||||
The `Chat Agent` could only respond via Emacs buffer insertion, rendering it useless for external channels like Telegram.
|
||||
** Root Cause
|
||||
Architectural myopia. The original chat skill assumed the user was always in front of Emacs.
|
||||
** Resolution
|
||||
Refactored `org-skill-chat` to be **Channel-Aware**:
|
||||
- It now extracts `:channel` and `:chat-id` from the inbound stimulus.
|
||||
- It dynamically generates the Probabilistic Engine mandate, instructing the LLM to use the appropriate `:target` (e.g., `:telegram`) based on the conversation context.
|
||||
|
||||
* 3. Side-Issue: UIOP Portability
|
||||
** Symptoms
|
||||
Tests failed with `Symbol "SETENV" not found in the UIOP/DRIVER package`.
|
||||
** Root Cause
|
||||
Misinterpretation of the `UIOP` API. `setenv` is not a standard export; the portable way is using `(setf (uiop:getenv ...) ...)`.
|
||||
** Resolution
|
||||
Updated all test environment setup to use the `setf` accessor.
|
||||
|
||||
* 4. opencortex Mandate Alignment
|
||||
** Autonomous Boundary
|
||||
By moving the Telegram API logic to a user-space skill and communicating with the core via standard stimuli, we have respected the microkernel boundary.
|
||||
** Homoiconic Memory
|
||||
All Telegram interactions are now logged as `:chat-message` events, ensuring the agent's history is unified regardless of the platform.
|
||||
|
||||
* 5. Permanent Learnings
|
||||
- **Foundation First:** Registries and logging macros must reside in the most foundational layers (`protocol` or `package`) to avoid load-order fragility.
|
||||
- **Instruct the Actuator:** When adding new channels, always update the Chat Agent's neural prompt so it knows how to "speak" back through the new interface.
|
||||
@@ -1,30 +0,0 @@
|
||||
#+TITLE: Root Cause Analysis: Containerized Infrastructure (Docker)
|
||||
#+DATE: 2026-04-11
|
||||
#+FILETAGS: :rca:docker:deployment:infrastructure:autonomy:
|
||||
|
||||
* Executive Summary
|
||||
Standardized the `opencortex` execution environment by creating a production-grade Docker infrastructure. This ensures that all system dependencies, including the Lisp runtime and external binaries like `signal-cli`, are locked down and portable.
|
||||
|
||||
* 1. Architectural Intent: The "Clean Room" Model
|
||||
** Problem
|
||||
The `opencortex` was relying on host-local binaries (`sbcl`, `signal-cli`) and manually configured Quicklisp dists. This made deployment to other environments (e.g., a VPS or a Autonomous Home Server) fragile and prone to version drift.
|
||||
** Solution
|
||||
1. **Dockerfile:** Created a multi-step build process that installs Debian Bookworm, SBCL, Java, and `signal-cli 0.14.0`.
|
||||
2. **Pre-Caching:** The build process triggers a `ql:quickload` of the `:opencortex` system, ensuring all Lisp dependencies are pre-downloaded and stored in the image layer, drastically reducing startup time.
|
||||
3. **Compose Orchestration:** Standardized the runtime via `docker-compose.yml`, which handles volume mounting of the user's `memex` directory and injection of `.env` secrets.
|
||||
|
||||
* 2. Volume Mapping & Persistence
|
||||
** Strategy
|
||||
To maintain the "Autonomous" mandate, the agent's code is isolated, but its memory (the `memex`) remains on the host.
|
||||
- **Mapping:** `../..` (host) -> `/memex` (container).
|
||||
- **State:** Created a named Docker volume `signal-state` to ensure that `signal-cli` identities and cryptographic keys survive container restarts and image updates.
|
||||
|
||||
* 3. Alignment with opencortex Mandates
|
||||
** Evolutionary Completion
|
||||
By moving to Docker, we have achieved "Evolutionary Completion" for the deployment track. The system is no longer a collection of scripts; it is a deployable appliance.
|
||||
** Documentation
|
||||
A new `Deployment Guide` was added to `docs/deployment.org` to ensure standard operating procedures are preserved.
|
||||
|
||||
* 4. Permanent Learnings
|
||||
- **Lisp Build Layers:** Always push the system to the ASDF registry and quickload during Docker build to bake dependencies into the image.
|
||||
- **Compose Locality:** Placing the `docker-compose.yml` inside the `projects/opencortex/` folder keeps infrastructure code close to the implementation logic.
|
||||
@@ -1,33 +0,0 @@
|
||||
#+TITLE: Root Cause Analysis: Asynchronous Lisp Repair Syntax Gate
|
||||
#+DATE: 2026-04-11
|
||||
#+FILETAGS: :rca:lisp:repair:decoupling:architecture:autonomy:
|
||||
|
||||
* Executive Summary
|
||||
Reimplemented the `org-skill-lisp-repair` to align with the "Autonomous Boundary" mandate. The previously synchronous, core-blocking repair logic has been replaced with an asynchronous, event-driven architecture using the Reactive Signal Pipeline.
|
||||
|
||||
* 1. Issue: Core Bloat & Synchronous Coupling
|
||||
** Symptoms
|
||||
The initial implementation of the Lisp Repair gate placed a `handler-case` and a dynamic function call (`repair-lisp-syntax`) directly inside the core `think` function (`probabilistic.lisp`). This forced the core to wait for repairs and made it "aware" of specific repair logic.
|
||||
** Root Cause
|
||||
Architectural shortcutting. By placing repair logic in the core execution path, we violated the microkernel principle which mandates that the core should be a "dumb" signal processor.
|
||||
** Resolution
|
||||
1. **Refactored Core:** `think` now only emits a `:syntax-error` stimulus if parsing fails. It no longer attempts to repair.
|
||||
2. **Asynchronous Skill:** `skill-lisp-repair` now triggers on the `:syntax-error` event. It performs the repair and returns the corrected action, which is then dispatched by the pipeline.
|
||||
|
||||
* 2. Side-Issue: Nested Signal Payloads
|
||||
** Symptoms
|
||||
`TYPE-ERROR` during testing when extracting the broken code from the stimulus.
|
||||
** Root Cause
|
||||
Mismatched expectations of signal nesting. The skill expected the code at `(getf context :payload)`, but in the `decide-gate`, `context` is the full signal, and the error details were nested inside the `:candidate` field of that signal.
|
||||
** Resolution
|
||||
Updated the deterministic logic to correctly traverse the nested signal structure: `(getf (getf context :candidate) :payload)`.
|
||||
|
||||
* 3. opencortex Mandate Alignment
|
||||
** Autonomous Boundary
|
||||
The core is now strictly a parser. Repair is an optional, user-space service.
|
||||
** Reactive Signal Pipeline
|
||||
Leveraged the pipeline's ability to re-inject `EVENT` signals to flatten the recursion of the repair loop.
|
||||
|
||||
* 4. Permanent Learnings
|
||||
- **Emit, Don't Call:** In a microkernel, if a non-fatal error occurs, always emit a signal rather than calling a recovery function. This allows the system to remain asynchronous and modular.
|
||||
- **Signal Inspection:** When writing deterministic gates, always verify the exact shape of the `context` signal being passed by the harness to avoid nesting errors.
|
||||
@@ -1,39 +0,0 @@
|
||||
#+TITLE: Root Cause Analysis: Playwright-Python Bridge (High-Fidelity Browsing)
|
||||
#+DATE: 2026-04-11
|
||||
#+FILETAGS: :rca:intelligence:browsing:automation:autonomy:
|
||||
|
||||
* Executive Summary
|
||||
Successfully implemented a high-fidelity browsing bridge using Playwright and Python. This allows the `opencortex` to interact with modern, JavaScript-rendered web applications that were previously inaccessible via simple HTTP clients.
|
||||
|
||||
* 1. Architectural Strategy: The I/O Bridge
|
||||
** Problem
|
||||
Common Lisp lacks a mature, native Playwright implementation. Direct bindings are complex and fragile.
|
||||
** Resolution
|
||||
Implemented a **JSON-over-STDIO Bridge**.
|
||||
- A standalone Python script (`browser-bridge.py`) manages the Playwright lifecycle and Chromium instance.
|
||||
- The Lisp kernel communicates with this script using `uiop:run-program`, passing parameters via `stdin` and receiving structured results via `stdout`. This provides a stable, decoupled interface.
|
||||
|
||||
* 2. Environment & Dependency Management
|
||||
** Issue
|
||||
Playwright requires a specific version of Chromium and several system-level libraries not present in the base Debian image.
|
||||
** Resolution
|
||||
Updated the `Dockerfile` to:
|
||||
1. Install Python3, pip, and venv.
|
||||
2. Create a virtual environment for isolated dependency management.
|
||||
3. Install the `playwright` package and execute `playwright install --with-deps chromium` during the image build. This ensures the production container is ready for high-fidelity browsing immediately upon startup.
|
||||
|
||||
* 3. Cognitive Tooling
|
||||
Created the `:browser` cognitive tool, which exposes three primary capabilities to Probabilistic Engine:
|
||||
- **Navigation:** Full JS rendering and waiting for network idle.
|
||||
- **Extraction:** Targeted text retrieval via CSS selectors.
|
||||
- **Vision:** Base64-encoded screenshot capture for future multimodal processing.
|
||||
|
||||
* 4. opencortex Mandate Alignment
|
||||
** Zero-Bloat (Managed)
|
||||
While adding Playwright increases the image size, it is a "Complexity Earned" trade-off that dramatically expands the agent's capability frontier.
|
||||
** Literate Granularity
|
||||
The `org-skill-playwright.org` file strictly follows the "one definition per block" mandate.
|
||||
|
||||
* 5. Permanent Learnings
|
||||
- **Inter-Process JSON:** JSON is the ideal lingua franca for Lisp-Python bridges.
|
||||
- **Path Portability:** Always use `uiop:native-namestring` when passing Lisp paths to external shell commands to ensure OS compatibility.
|
||||
@@ -1,40 +0,0 @@
|
||||
#+TITLE: Root Cause Analysis: Individual Provider Track Verification
|
||||
#+DATE: 2026-04-11
|
||||
#+FILETAGS: :rca:providers:llm:testing:autonomy:
|
||||
|
||||
* Executive Summary
|
||||
Verified the unified LLM gateway implementation for all 6 individual provider tracks (Anthropic, Gemini, Groq, OpenAI, OpenRouter, Ollama). Identified and resolved critical parsing failures in the Gemini track and integration gaps in the system build definition.
|
||||
|
||||
* 1. Issue: Fragile Response Parsing (Gemini)
|
||||
** Symptoms
|
||||
Gemini API responses were returning `NIL` content during mocked unit tests, despite the JSON structure being seemingly correct.
|
||||
** Root Cause
|
||||
Recursive `assoc` / `car` / `cdr` chains were hardcoded and brittle. Specifically, the Gemini extraction logic was incorrectly attempting to treat a single alist pair as a list of pairs, causing `assoc` to fail on the `:TEXT` key.
|
||||
** Resolution
|
||||
Implemented a robust `get-nested` helper function that safely traverses both nested objects (alists) and arrays (lists of alists). This normalized the extraction logic across all providers.
|
||||
|
||||
* 2. Issue: Decoupled Build Configuration
|
||||
** Symptoms
|
||||
Provider logic was present in the codebase but inaccessible during tests and runtime.
|
||||
** Root Cause
|
||||
The `credentials-vault.lisp` and `llm-gateway.lisp` files (consolidated in a previous session) were never added to the `opencortex.asd` system definition. Furthermore, an incorrect loading order caused `UNDEFINED-FUNCTION` errors for `register-probabilistic-backend`.
|
||||
** Resolution
|
||||
1. Added both files to `opencortex.asd`.
|
||||
2. Enforced strict loading order: `probabilistic` (defines registry) -> `credentials-vault` -> `llm-gateway` (uses registry).
|
||||
|
||||
* 3. Issue: Credential Key Mismatch
|
||||
** Symptoms
|
||||
Gemini requests failed with "API Key missing" even when environment variables were set.
|
||||
** Root Cause
|
||||
`llm-gateway` requested secrets for the `:gemini-api` provider, but the `credentials-vault` fallback logic only recognized the `:gemini` keyword.
|
||||
** Resolution
|
||||
Updated `vault-get-secret` to map both `:gemini` and `:gemini-api` to the same `GEMINI_API_KEY` environment variable.
|
||||
|
||||
* 4. opencortex Mandate Alignment
|
||||
** Invariant Check
|
||||
- *High-Integrity Memory:* All individual provider tracks are now backed by automated unit tests (`llm-gateway-tests.lisp`).
|
||||
- *Literate Programming:* Updated `org-skill-llm-gateway.org` to reflect the improved `get-nested` utility.
|
||||
|
||||
* 5. Permanent Learnings
|
||||
- **Tooling vs Source:** Tangled `.lisp` files are not enough; always ensure new modules are registered in the `.asd` file to be part of the official kernel build.
|
||||
- **Robustness over Brevity:** Use abstraction helpers like `get-nested` instead of deep `car/cdr` chains when dealing with external JSON structures that may have varying array/object nesting.
|
||||
@@ -1,40 +0,0 @@
|
||||
#+TITLE: Root Cause Analysis: Autonomous Self-Fix Loop Verification
|
||||
#+DATE: 2026-04-11
|
||||
#+FILETAGS: :rca:self-fix:autonomy:testing:
|
||||
|
||||
* Executive Summary
|
||||
Verified the autonomous repair capability of the `Self-Fix Agent`. The system successfully detected a deterministic type error in a secondary skill, initiated a repair request, and programmatically patched the source code via the `:repair-file` tool.
|
||||
|
||||
* 1. Issue: Self-Fix Mechanism Verification
|
||||
** Symptoms
|
||||
Manual verification was required to prove that `org-skill-self-fix` could transition from "Thinking" about a bug to "Acting" on the file system.
|
||||
** Root Cause
|
||||
N/A (Deterministic test injection).
|
||||
** Resolution
|
||||
Created `self-fix-tests.lisp` which:
|
||||
1. Generates `org-skill-broken-math.org` with a `(+ 1 "two")` bug.
|
||||
2. Triggers the bug to produce a `PIPELINE CRASH`.
|
||||
3. Injects a `:repair-request` stimulus.
|
||||
4. Executes `self-fix-apply` to replace the bug with `(+ 1 2)`.
|
||||
5. Verifies the file content and successful hot-reload.
|
||||
|
||||
* 2. Side-Issue: ASDF Configuration Fragility
|
||||
** Symptoms
|
||||
Repeated `LOAD-SYSTEM-DEFINITION-ERROR` and "unmatched close parenthesis" errors during test integration.
|
||||
** Root Cause
|
||||
Complexity in the `:components` nesting of `opencortex.asd` led to repeated syntax errors when using automated editing tools. The deep nesting made manual paren counting prone to "off-by-one" errors.
|
||||
** Resolution
|
||||
Refactored `opencortex.asd` to use a **Flat Component Structure**.
|
||||
- *Before:* `:components ((:module "src" :components (...)))`
|
||||
- *After:* `:components ((:file "src/package") ...)`
|
||||
This eliminates unnecessary nesting levels and drastically reduces the surface area for syntax errors.
|
||||
|
||||
* 3. opencortex Mandate Alignment
|
||||
** Invariant Check
|
||||
- *Lisp Machine Autonomousty:* Verification utilized hot-reloading (`load-skill-from-org`) without restarting the SBCL image.
|
||||
- *Literate Programming:* Updated `org-skill-self-fix.org` to match the finalized `self-fix.lisp` logic.
|
||||
- *Institutional Memory:* This RCA documents the decision to flatten the `.asd` structure to prevent future "Parenthesis Hell" incidents.
|
||||
|
||||
* 4. Permanent Learnings
|
||||
- **Flatten Configuration:** Keep `defsystem` definitions as flat as possible. The overhead of `:module` blocks often outweighs their organizational benefit in a probabilistic-deterministic environment where agents frequently edit these files.
|
||||
- **Mocking Probabilistic Engine:** For verifying *loop mechanics*, mocking LLM responses is essential to ensure test determinism, while integration tests can use live LLM calls.
|
||||
@@ -1,33 +0,0 @@
|
||||
#+TITLE: Root Cause Analysis: Shell Actuator Security Hardening
|
||||
#+DATE: 2026-04-11
|
||||
#+FILETAGS: :rca:security:shell:injection:autonomy:
|
||||
|
||||
* Executive Summary
|
||||
During the formal verification of the `org-skill-shell-actuator`, a critical command injection vulnerability was identified and patched. The previous implementation relied on a naive whitelist check that could be bypassed using shell metacharacters.
|
||||
|
||||
* 1. Issue: Command Injection Vulnerability
|
||||
** Symptoms
|
||||
Commands like `ls ; rm -rf /` were potentially executable if the first word (`ls`) was in the whitelist.
|
||||
** Root Cause
|
||||
The `execute-shell-safely` function only checked the first space-delimited word of the command string against the `*allowed-commands*` whitelist. Since `uiop:run-program` executes string-based commands via `/bin/sh -c`, the shell would process the entire string, including injected commands following metacharacters like `;`, `&`, or `|`.
|
||||
** Resolution
|
||||
1. **Metacharacter Blacklist:** Introduced `*shell-metacharacters*` containing dangerous shell symbols (`; & | > < $ \` \ !`).
|
||||
2. **Strict Validation:** Updated `execute-shell-safely` to scan the *entire* command string for these characters before performing the whitelist check.
|
||||
3. **Defense-in-Depth:** Any command containing a metacharacter is now rejected with a "Security Violation" error, even if the primary command is whitelisted.
|
||||
|
||||
* 2. Side-Issue: Missing Package Context
|
||||
** Symptoms
|
||||
`UNDEFINED-FUNCTION EXECUTE-SHELL-SAFELY` during unit tests.
|
||||
** Root Cause
|
||||
`src/shell-logic.lisp` was missing an `(in-package :opencortex)` declaration, causing symbols to be defined in the default `COMMON-LISP-USER` package instead of the harness package.
|
||||
** Resolution
|
||||
Added the `in-package` header to `shell-logic.lisp`.
|
||||
|
||||
* 3. opencortex Mandate Alignment
|
||||
** Invariant Check
|
||||
- *High-Integrity Memory:* The shell actuator is now formally verified with 4 new unit tests covering whitelist enforcement and injection blocking.
|
||||
- *Literate Programming:* Updated `org-skill-shell-actuator.org` Phase A and Build sections to reflect the hardened logic.
|
||||
|
||||
* 4. Permanent Learnings
|
||||
- **Whole-String Validation:** Never assume that whitelisting the "head" of a command string is sufficient when passing that string to a shell.
|
||||
- **Subshell Avoidance:** While the current fix blacklists metacharacters, future iterations should move toward passing command arguments as a Lisp list to `uiop:run-program`, bypassing the shell entirely.
|
||||
@@ -1,48 +0,0 @@
|
||||
#+TITLE: Root Cause Analysis: Consolidation VI - Task Orchestrator Implementation
|
||||
#+DATE: 2026-04-11
|
||||
#+FILETAGS: :rca:orchestrator:consensus:integrity:
|
||||
|
||||
* Executive Summary
|
||||
The implementation of Consolidation VI (Task Orchestrator) aimed to introduce parallel multi-backend consensus, GTD task integrity, and delegation. During the build, a critical dependency failure was identified in the `lisp-validator` module.
|
||||
|
||||
* 1. Issue: Undefined `SAFETY-HARNESS-VALIDATE`
|
||||
** Symptoms
|
||||
Existing `SAFETY-SUITE` tests failed with `#<UNDEFINED-FUNCTION SAFETY-HARNESS-VALIDATE>`.
|
||||
** Root Cause
|
||||
The function `lisp-validator-validate` was exported in `package.lisp` but never actually defined in `lisp-validator.lisp`. Only the internal recursive walker `lisp-validator-ast-walk` existed. This represents a "Hollow Export" bug where the interface was designed but the implementation was truncated or skipped in a previous session.
|
||||
** Resolution
|
||||
Defined `lisp-validator-validate` as a wrapper around `read-from-string` and `lisp-validator-ast-walk`.
|
||||
|
||||
* 2. Design Decision: Deterministic Consensus
|
||||
** Requirement
|
||||
Multi-backend support to reduce hallucinations and increase reliability.
|
||||
** Solution
|
||||
Implemented `bt:make-thread` parallel queries in `ask-probabilistic`.
|
||||
** Trade-off
|
||||
Selected "Majority Rules" over "First-to-Finish".
|
||||
- *Pros:* Higher accuracy, mathematically consistent.
|
||||
- *Cons:* Slower (latency limited by the slowest provider).
|
||||
** Invariant Alignment
|
||||
Aligns with opencortex Mandate 4 (Radical Transparency) and Invariant 2 (Technical Mastery) by ensuring decisions are auditable and consistent across multiple brains.
|
||||
|
||||
* 3. Design Decision: Task Integrity Gate
|
||||
** Requirement
|
||||
Prevent illegal GTD state transitions.
|
||||
** Solution
|
||||
Added `task-integrity-check` in `deterministic.lisp`.
|
||||
** Invariant Alignment
|
||||
Enforces the "High-Integrity Memory" mandate by ensuring the Org-mode AST remains semantically valid according to GTD rules (e.g., no orphaned active tasks).
|
||||
|
||||
* 4. opencortex Mandate Violations during Session (Corrected)
|
||||
** Violations
|
||||
1. Editing without prior commit.
|
||||
2. Direct `.lisp` edits vs Literate Org tangling.
|
||||
3. Multi-function edits per block.
|
||||
** Correction
|
||||
1. Performed a retrospective commit.
|
||||
2. Synchronized `probabilistic-deterministic.org` and `core.org` with source code.
|
||||
3. Refactored the Markdown flight plan into an Org-mode flight plan.
|
||||
|
||||
* 5. Permanent Learnings
|
||||
- *Check Exports:* Always verify that symbols exported in `package.lisp` have a corresponding definition in the literate source.
|
||||
- *Strict opencortex Mode:* Enable a pre-save hook or agent check to ensure all edits are performed within `#+begin_src` blocks in Literate Org files to avoid synchronization debt.
|
||||
Reference in New Issue
Block a user