@matheuskrumenauer/tanya 0.4.0-beta.0 → 0.5.0-beta.0

package/README.md

@@ -163,7 +163,7 @@ through the permission engine before treating them as safe extension points.
 
 `--verify` adds required verification commands to the run context. Tanya must run and report each exact command before finishing the coding task.
 
-`tanya benchmark run --all` currently exercises 26 executable low-to-medium regression fixtures: targeted edits, new files, dependency/lockfile updates, framework-style migrations, failing-test repair, frontend smoke checks, artifact/context reuse, streaming long-tool execution, run-history logging, dirty worktrees, report repair, and the CosmoHQ mobile/backend smoke profiles.
+`tanya benchmark run --all` currently exercises 27 executable low-to-medium regression fixtures: targeted edits, new files, dependency/lockfile updates, framework-style migrations, failing-test repair, frontend smoke checks, artifact/context reuse, streaming long-tool execution, compaction-boundary recovery, run-history logging, dirty worktrees, report repair, and the CosmoHQ mobile/backend smoke profiles.
 
 By default, `tanya run` also performs an independent post-check after the agent finishes. If the workspace has a `typecheck` script, Tanya reruns that exact script with the local package manager (`npm`, `pnpm`, `yarn`, or `bun`). If not, it falls back to `npx tsc --noEmit --pretty false` when a `tsconfig` is present. If the workspace has a `test` script, Tanya reruns that as well unless the run already reported a passing test verification.
 
@@ -201,6 +201,19 @@ Long-running `run_shell` calls stream throttled stdout/stderr chunks to the acti
 tanya run "Run npm test"  # emits tool_progress while the command runs; Ctrl-C cancels the active shell and returns partial_output
 ```
 
+## Long sessions
+
+Tanya handles context pressure as a cascade instead of truncating abruptly:
+
+1. Microcompact folds empty/no-op tool-call pairs in place.
+2. Snip removes low-signal history such as duplicate file reads and empty read-only tool results.
+3. Auto-compact reacts to provider `413` / context-window errors by summarizing older turns into a `[compaction summary: ...]` system message and retrying once normally, then once more aggressively.
+4. Archive writes compacted messages to `.tanya/runs/<runId>/archive.jsonl` before they leave live history, so verifier scans and future memory tools can still inspect them.
+
+Runs are capped at three total auto-compactions. If the provider still rejects the context, Tanya raises `CompactionExhaustedError` and asks the user to narrow the task, clear the session, or split the work.
+
+See [docs/long-sessions.md](./docs/long-sessions.md) for details.
+
 Context files are generic JSON envelopes for caller-supplied task metadata, artifacts, instructions, and verification commands.
 
 ## Current Tools