Skip to main content
Released September 4, 202624 new features · 34 fixesGitHub releaseCompare changes

New features

  • audit record every LLM request on every surface in the audit trail #1448 · a4a6fac
  • cache explicit Gemini cache resources, storage pricing and auto TTL #1451 · b719cfa
  • cache rolling conversation breakpoint, extended TTL and cache telemetry across providers #1442 · 4ec16b3
  • compaction session threshold, dedicated summarizer, tool-catalog reserve and repeated-read dedup #1445 · 144e02c
  • config organization-managed defaults and locked policies #1456 · e360582
  • context compaction hooks, verbatim-safe levels, summarizer accounting, 1h cache TTL on Bedrock/OpenRouter #1460 · d1e77cc
  • context incremental re-index with file stamps, /context refresh and /context watch #1455 · c969c28
  • context live context inspector and learned chars-per-token calibration #1443 · fa0dfb2
  • context persist the token calibrator per state root and route every estimate through it #1462 · 6c22c4b
  • context prefix budget allocator, projected ctx%, /clear conversation semantics #1453 · 445263d
  • gateway per-principal store sets under hub isolation #1450 · 6b46d74
  • gateway tenant-scoped park and task-graph roots and a per-tenant daily budget #1467 · 7ead7fc
  • knowledge opt-in lexical normalization, fact IDF, local embeddings with cost, retrieval budget and citations #1465 · 2a40a7b
  • memory export/import, multi-process merge for profile/topics/projects, recall reasons and project labels #1464 · a2b9b95
  • memory pluggable memory provider and context engine through MCP #1457 · 8adc9c5
  • memory vector-aware auto-recall with a relevance floor and session-persisted context attachments #1447 · 311cf1a
  • retention expire park and cost snapshots on the session TTL and show every store’s policy #1449 · 09a4a6a
  • retrieval rerank stage, corpus weights, structure-aware passages and evidence-based recall #1452 · 41b267b
  • security encryption at rest for memory, contexts, CCR and costs, key rotation and a hash-chained audit trail #1466 · 442b2e6
  • session append-only transcript journal and memory flush before compaction #1446 · 0c34ce0
  • session undo compaction, persisted rewind checkpoints and transcript export/search #1461 · 909f050
  • telemetry OpenTelemetry metrics export over OTLP/HTTP and a provider-side context engine #1468 · 6a0e12b
  • tokens provider token counting on every surface, exact calibration and budgeted summaries #1454 · 19d57ac
  • workspace nested AGENTS.md hierarchy with imports, nested watch dirs and legacy source paths #1463 · 607a41c

Fixes

  • agent follow the @model route in delegation and never report an empty worker as success #1440 · c965332
  • agents workers share the orchestrator’s compactor, overflow recovery and journal #1490 · ef08a6b
  • cache provider cache parity — Bedrock system marker TTL, GPT-5.5 retention, cache key on OpenAI-compatible upstreams, OpenRouter marker cap #1479 · 1545c42
  • claudeai surface SSE error events and decode br/zstd responses #1458 · edc1148
  • config registry, reload and surface truth — missing defaults, reloadable variables, /config memory rows, /help lines, completer entries, sections hint from the router, dead code #1494 · 64a50d5
  • context byte-stable system prompt with per-turn context as a flagged user message, stable cache key, frozen prefix ratio #1470 · 136526d
  • context cache-stable prompt prefix, reversible guided compact and fsync durability #1438 · 339cd79
  • context close the P0 gaps of the second context audit #1459 · a026988
  • context mirror provider context edits locally — stub cleared tool results, skip calibration, note the rebuild, surface freed tokens #1483 · cd30482
  • context recover from context overflow on chat, RPC, one-shot and MoA; classify every provider’s overflow body; per-session recovery budgets #1477 · fd84ffe
  • context single context estimate and window-scaled retrieval budget #1485 · 0d00f9c
  • context verbatim results survive pairing, verbatim cap for convergence, skipped-compaction bookkeeping, shared restore helper #1472 · 821638c
  • cost attribute usage to the call that produced it — summarizer, memory worker, streaming, Gemini thinking, provider resets #1473 · fc96fa9
  • cost cache telemetry truth — subset-schema total misses, per-provider first and hit ratio, expected rebuilds on every prefix change, Bedrock TTL as sent #1481 · a11449d
  • cost long-context tiers per call, Amazon Nova rows, Copilot as a subscription #1491 · ca39638
  • cost reset clears every aggregate, daily-aware budget notices, summarizer cache keyed by env, locked audit path fails closed #1488 · c4e8103
  • cost telemetry semconv and opt-in session series, embedding and daily-budget metrics, CSV export, spend finalized on every surface, park on hard stop #1496 · 78106ac
  • gateway swap every per-conversation state with the tenant, guard the base memory worker, tag external memory calls, keep the workspace on queued segments #1474 · 42695ff
  • knowledge batched, single-flight embedding with skip-and-continue; binary/minified skip; passage hard cap; partial upsert persistence; rune-safe cuts #1478 · c5946f2
  • knowledge hybrid retrieval as a vector-BM25 union fused by RRF, query embedded once per turn, system-only tagged reranker #1486 · 381bfdd
  • memory cross-process lock around every merge-and-write of the shared stores #1489 · 0b3ae15
  • memory drain the queued backlog at boot and on every tick, bounded shutdown wait, exit-tail queueing, race-free watermark #1480 · b156159
  • memory fence recalled and retrieved text as data, keep injected blocks out of extraction and search, sanitize and validate stored facts #1482 · a95c760
  • memory hygiene batch — rune-safe cuts, test-only helpers out of production, CCR skip surfaced, coder read line cap, transcript cap and plaintext notice, isolation refusal, external-only memory provider, worktree-aware projects #1495 · 75b8f98
  • memory lock sealed stores read-only when the key is missing, keep daily notes plain, seal and redact the pending queue #1469 · 8065549
  • security at-rest v2 — payloads bound to their store and tenant, passphrases stretched, fsync on reseal, 16-byte tenant digest, 0700 dirs #1493 · 47a3f50
  • security encrypt session stores at rest and redact secrets on the LLM path #1441 · a90e35b
  • security never hand ChatCLI’s own secrets to MCP servers and executable plugins #1475 · 495a713
  • security persisted-data hygiene — strict-policy redaction on every store, sealed REPL history, missing secret shapes, periodic tenant retention #1484 · ab1f286
  • security share one locked, rotating, sealable audit chain across the REPL, gateway and gRPC server #1476 · 39c469d
  • session exact schema gate, fork on its own journal with attachments, cost and CCR references, atomic state writes #1487 · b3a5110
  • session journal integrity — full-prefix hash scan, tool-call-aware hashes, torn-line tolerance, no tail dedup #1471 · d52634f
  • ui footer shows used context apart from the answer reserve #1497 · 14df167
  • workspace instruction-file parity — import boundary and per-file cap, global always merged, local and override files, .claude/rules, watched-dir cap with .gitignore #1492 · f75aaf1

Newer release: v1.196.1

a reload keeps what the client provided and reaches the servers; environment file discovery for editor-spawned servers and Bedrock profile parity

Previous release: v1.195.0

list the models the account can invoke via the Devin CLI