fix(fs-local): degrade basis I/O failures to null

Descriptor-phase errnos in readTextForDiff fold to before: null so a file deleted or made unreadable after the caller's preflight cannot fail the committed write; cancellation and non-errno faults still propagate. Drops the now-covered isFile v8 ignore, extends llm-replay with catalog capability parity (defaultMaxTokens/reasoningEfforts), and records the fs-write-overwrite-bounded keyless snapshot pinning the over-limit whole-file fallback through the real acp-agent composition.
2026-08-08 11:49:42 -07:00
parent 3f3c8e31e9
commit 9a299f9827
13 changed files with 281 additions and 43 deletions
@@ -0,0 +1,46 @@
+# Keyless replay counterpart to fs-diff-bound.cordis.yml. Replay patches apply
+# directly against the live cordis.yml because include patches cannot target
+# entries behind a nested include; the acp-agent restatement keeps the recorded
+# deepseek-v4-flash model and raw JSONL persistence for the harness's harvest.
+- id: base
+  name: '@cordisjs/plugin-include'
+  config:
+    path: ../cordis.yml
+    patches:
+      - id: llm-deepseek
+        name: '@deepseek-ai/dsh-llm-deepseek'
+        disabled: true
+      - id: acp-agent
+        name: '@deepseek-ai/dsh-acp-demo'
+        config:
+          provider: deepseek-official
+          model: deepseek-v4-flash
+          persistenceRoot: !!js process.env.DSH_SNAPSHOT_SESSIONS_ROOT ?? './.sessions'
+          persistenceCompression: none
+          workspaceContext:
+            maxBytes: 65536
+          persona: |
+            You are a coding assistant powered by the {{model}} model. Your working directory is {{cwd}}. Your bash tool runs under a file sandbox — a `[sandbox: file access denied …]` result is policy, not a command bug.
+
+            Verify your work by running the code or tests. Keep answers brief and factual.
+      - id: fs-sandbox
+        name: '@deepseek-ai/dsh-fs-sandbox'
+        config:
+          cwd: !!js process.cwd()
+          diffBasisMaxBytes: 64
+      - insert:
+          - id: llm-replay
+            name: '@deepseek-ai/dsh-llm-replay'
+            config:
+              providers:
+                - id: deepseek-official
+                  name: DeepSeek
+                  models:
+                    # Capability parity with the live adapter so replay
+                    # reconstructs the freshly recorded request header.
+                    - id: deepseek-v4-flash
+                      contextWindow: 1000000
+                      defaultMaxTokens: 256000
+                      reasoningEfforts: ['off', 'high', 'max']
+                      defaultReasoningEffort: max
+                    - id: deepseek-v4-pro