The research tool now spawns the scout agent in a background tokio task
and returns immediately with a research_id placeholder. This allows the
agent to continue working while research runs (30-120 seconds).
Key changes:
- New PendingResearchManager for tracking async research tasks
- research tool returns immediately with placeholder containing research_id
- research_status tool to check progress of pending research
- Auto-injection of completed research at natural break points:
- Start of each tool iteration (before LLM call)
- Before prompting user in interactive mode
- /research CLI command to list all research tasks
- Updated system prompt to explain async behavior
The agent can:
- Continue with other work while research runs
- Check status with research_status tool
- Yield turn to user if results are critical before continuing
- Fix test_rehydrate_success race condition by using UUID for unique session IDs
- Add #[serial] attribute to prevent parallel execution conflicts
- Improve cleanup to remove entire session directory tree
- Add characterization test for resize_image_to_dimensions fallback behavior
(documents fix from commit af8b849 for media type preservation)
Agent: hopper
When resize_image_to_dimensions() returns a larger file than the original,
we fall back to using the original bytes. Previously, was_resized was set
to true if the original dimensions exceeded MAX_IMAGE_DIMENSION, which
caused final_media_type to be set to 'image/jpeg' even though we were
using the original PNG bytes.
This caused Anthropic API errors like:
'Image does not match the provided media type image/jpeg'
Fix: Set was_resized=false when falling back to original bytes, so the
original media type (detected from magic bytes) is preserved.
Rename all references from "Project Memory" to "Workspace Memory" to avoid
future conflation if a "project" concept is introduced later.
Changes:
- Rename read_project_memory() -> read_workspace_memory()
- Update all prompts, tool descriptions, and comments
- Update header parsing in memory.rs to use "# Workspace Memory"
- Update display detection for "=== Workspace Memory ==="
- Update documentation and analysis/memory.md
11 files changed, ~36 occurrences updated.
The Anthropic API was rejecting requests with multiple high-resolution images
(~2000x3000 pixels each) even though individual file sizes were under limits.
Root cause: Code only checked per-image file size (3.75MB), not dimensions.
Claude recommends images ≤1568px on longest edge and has 32MB total request limit.
Changes:
- Add MAX_IMAGE_DIMENSION (1568px) and MAX_TOTAL_IMAGE_PAYLOAD (20MB) constants
- Trigger resize when dimensions > 1568px (not just file size > 3.75MB)
- Add new resize_image_to_dimensions() for dimension-constrained resizing
- Track cumulative payload size across multiple images
- Warn if total payload exceeds recommended limit
Test results with Walking Dead comic images:
- WD_0001_0001.jpg: 800KB 1987x3057 → 321KB 1019x1568
- WD_0001_1064.png: 150KB 1988x3057 → 143KB 1020x1568
- WD_0002_0001.jpg: 1023KB 1988x3056 → 292KB 1020x1568
- Total payload: ~2.5MB → ~1MB base64
Removed the persistent_chrome config flag - chromedriver is now always
kept running after webdriver_quit. This eliminates startup latency for
subsequent WebDriver sessions.
Safaridriver is still killed on quit since it doesn't benefit from
persistence in the same way.
Updated quit message to correctly indicate chromedriver remains running.
When webdriver_start is called, now checks if chromedriver is already
running on the configured port and reuses it instead of spawning a new
process. This significantly reduces startup time for subsequent sessions.
New config option:
[webdriver]
persistent_chrome = true # Keep chromedriver running between sessions
When enabled, webdriver_quit closes the browser session but leaves
chromedriver running for reuse by the next session.
The Anthropic API has a 5MB limit on base64-encoded images, not raw file
size. Base64 encoding increases size by ~33% (4/3 ratio), so a 4MB raw
image becomes ~5.3MB encoded, exceeding the limit.
Changed MAX_IMAGE_SIZE from 5MB to ~3.75MB (5MB * 3/4) to trigger
resizing before the base64-encoded result exceeds the API limit.
Also updated target resize size to 3.6MB to leave margin.
Images >= 5MB are now automatically resized to < 4.9MB using ImageMagick
before being sent to the LLM. This prevents API errors from oversized images.
- Uses iterative quality/scale reduction to find optimal size
- Converts to JPEG for better compression
- Shows original and resized size in terminal output (e.g., '6.2 MB → 4.1 MB (resized)')
- Falls back to original if ImageMagick fails or isn't available
- Shell outputs > 8KB are truncated to first 500 chars
- Full output saved to .g3/sessions/<session_id>/tools/shell_stdout_<id>.txt
- LLM can use read_file with start/end to paginate through large outputs
- read_file now uses seek() for O(1) random access instead of reading entire file
- UTF-8 safe: reads extra bytes at boundaries to find valid char positions
- Falls back to lossy conversion for binary files (no panics)
Files changed:
- paths.rs: get_tools_output_dir(), generate_short_id()
- shell.rs: truncate_large_output() integration
- file_ops.rs: seek-based read_file_range() helper
- New test: read_file_utf8_test.rs
Project memory is now stored at analysis/memory.md instead of .g3/memory.md.
This change enables:
- Shared memory across git worktrees (studio agent sessions)
- Version-controlled memory that persists across clones
- Memory changes tracked in git history and reviewable in PRs
Changes:
- crates/g3-core/src/tools/memory.rs: Update get_memory_path() to use analysis/
- crates/g3-cli/src/project_files.rs: Update read_project_memory() path
- crates/g3-core/src/prompts.rs: Update documentation references (2 occurrences)
- analysis/memory.md: Add memory file (copied from .g3/memory.md)
ACD (Aggressive Context Dehydration) fixes:
- Fixed dehydrate_context() to extract turn summary from context window
instead of using the passed-in final_response (which contained only
the timing footer, not the actual LLM response)
- Removed final_response parameter from dehydrate_context() since it
now self-extracts the last assistant message as the summary
- This ensures the actual turn summary is preserved after dehydration,
not just the timing footer
New /dump command:
- Added /dump command to dump entire context window to tmp/ for debugging
- Shows message index, role, kind, content length, and full content
- Available in both console and machine modes
UTF-8 safety:
- Fixed truncate_to_word_boundary() to use character indices instead of
byte indices, preventing panics on multi-byte UTF-8 characters
- Added UTF-8 string slicing guidance to AGENTS.md
Agent: g3
When read_file is called with an end position beyond the file length,
instead of returning an error that forces a retry, now clamps to the
actual file length and returns the content with an informative message.
This eliminates wasteful retry cycles where the LLM had to make a
second request with the corrected end position.
Change read_file output format so the "🔍 N lines read" appears as
the last line after the file content, not before it. This keeps the
output cleaner with just one metadata line at the end.
1. str_replace: Show insertion/deletion counts with colors
"✅ +N insertions | -M deletions" (green/red)
2. write_file: Compact format with human-readable sizes
"✅ wrote N lines | Xk chars"
3. read_file: Cleaner format
"🔍 N lines read" instead of "📄 File content (N lines)"
4. webdriver_quit: Show correct driver name (safaridriver vs chromedriver)
5. read_file: When start position exceeds file length, read last 100 chars
with explanation instead of failing
6. shell: Remove redundant "Command failed:" prefix from error messages
Collapsed nested if statements that check related conditions into
single conditions using &&. This improves readability by making
the logical relationship between conditions explicit.
Files changed:
- feedback_extraction.rs: 3 instances of tool_use/final_output checks
- tools/todo.rs: 1 instance of todo completion check
Agent: fowler
- Remove IMPORTANT FOR CODING section (~1,500 chars of coding guidelines)
- Remove <use_parallel_tool_calls> block (~500 chars)
- Remove unused const_format dependency from g3-core
- Simplify get_system_prompt_for_native() to just return base prompt
- Response Guidelines now cleanly ends the static prompt
Prompt reduced from ~8,500 to ~6,500 characters.
When the scout agent fails (e.g., context window exhaustion), now:
- Captures both stdout and stderr from the scout process
- Detects context window exhaustion errors with specific patterns
- Provides detailed, actionable error messages to the user
- Shows suggestions for how to work around the issue
- Includes technical details (exit code, error output) for debugging
Handles two failure modes:
1. Scout agent exits with non-zero status
2. Scout agent exits successfully but doesn't produce valid report markers
Both cases now surface clear error messages instead of cryptic failures.
The timing footer (e.g., ⏱️ 19.4s | 💭 4.7s) was being saved to the
conversation history as a separate assistant message. This happened
because stream_completion_with_tools returns the timing footer in
TaskResult.response for display, but the caller was also saving it
to context.
Fix: Strip the timing footer (identified by \n\n⏱️) before saving
to context window. The timing footer remains display-only.
Also includes:
- Research tool blank line fix: only add visual separator for research
tool output, not all tools
- Research tool webdriver propagation: pass parent's webdriver browser
choice (Safari vs Chrome headless) to scout subprocess
final_output removal:
- Remove final_output from tool definitions and dispatch
- Update system prompts to request summaries as regular text
- Remove final_output_called field from StreamingState
- Update auto_continue tests to remove final_output_called parameter
- Remove final_output test from tool_execution_test.rs
- Update planner and flock prompts to not reference final_output
- Keep backwards-compat code in feedback_extraction.rs and task_result.rs
Scout report handback:
- Change from file-based to delimiter-based report extraction
- Scout outputs report between ---SCOUT_REPORT_START/END--- markers
- Research tool extracts content between markers, strips ANSI codes
- Add comprehensive tests for extraction and ANSI stripping
657 tests pass.
The research tool now streams the underlying scout agent's output
to the CLI in real-time for visual indication of progress. This
output is displayed but not added to the conversation context.
New tool that spawns a scout agent to perform web research and return
a structured research brief. The scout agent uses webdriver to browse
the web and returns a decision-ready report.
Changes:
- Added 'research' tool definition (12 core tools total)
- Added research tool dispatch in tool_dispatch.rs
- Created tools/research.rs implementation:
- Spawns 'g3 --agent scout <query>' as subprocess
- Captures stdout and extracts last line (report file path)
- Reads and returns the report file contents
- Added exclude_research flag to ToolConfig
- Scout agent (agent_name == 'scout') does NOT have access to research
tool to prevent infinite recursion
- Updated system prompts to describe when to use research tool
- Added scout.md agent prompt with research brief output contract
The research tool is preferred for complex research tasks (APIs, SDKs,
libraries, approaches, bugs). WebDriver can still be used directly for
simple lookups or fine-grained control.
- Remove final_output from tool definitions, dispatch, and misc tools
- Update system prompts to request summaries as regular markdown text
- Remove print_final_output from UiWriter trait and all implementations
- Remove final_output handling from agent core logic
- Rename final_output_summary → summary in session continuation
- Delete final_output test files
- Update tool count tests (12→11, 27→26)
This allows LLM summaries to stream through the markdown formatter
for a more natural, responsive user experience instead of buffering
everything into a tool call.
This fixes a bug where the agent would stop responding abruptly without
calling final_output. The root cause was the allow_multiple_tool_calls
config option (default: false) which caused the agent to break out of
the streaming loop mid-stream after executing the first tool, losing
any subsequent content.
Changes:
- Remove allow_multiple_tool_calls config option entirely
- Always process all tool calls without breaking mid-stream
- Simplify system prompt generation (no longer needs boolean param)
- Let the stream complete fully before continuing to next iteration
- Change find_last_tool_call_start to find_first_tool_call_start
- Remove parser.reset() call on duplicate detection
Benefits:
- Simpler logic with less conditional branching
- No lost content after tool calls
- Consistent behavior for all users
- Reduced config complexity
Agent: carmack
Add get_session() helper function that:
- Checks if webdriver is enabled
- Acquires the session read lock
- Returns the cloned session or an error message
Refactored 12 webdriver tool functions to use this helper:
- execute_webdriver_navigate
- execute_webdriver_get_url
- execute_webdriver_get_title
- execute_webdriver_find_element
- execute_webdriver_find_elements
- execute_webdriver_click
- execute_webdriver_send_keys
- execute_webdriver_execute_script
- execute_webdriver_get_page_source
- execute_webdriver_screenshot
- execute_webdriver_back
- execute_webdriver_forward
- execute_webdriver_refresh
Each function previously had ~10 lines of identical boilerplate.
Now reduced to 4 lines using the helper.
Net reduction: 68 lines (678 -> 610)
All tests pass. Behavior unchanged.
Agent: carmack
databricks.rs:
- Extract ToolCallAccumulator struct to replace opaque (String, String, String) tuple
- Add decode_utf8_streaming() helper for cleaner UTF-8 handling
- Add is_incomplete_json_error() helper for JSON parse error detection
- Add make_final_chunk() helper to reduce duplication
- Add finalize_tool_calls() to convert accumulators to final format
- Refactor parse_streaming_response from ~270 lines to ~100 lines
- Reduce nesting depth from 8+ levels to 4 levels
- Use early returns and let-else for cleaner control flow
file_ops.rs:
- Replace repetitive if-let chains with declarative PATH_CONTENT_KEYS table
- Use match expression instead of nested if-else
- Reduce extract_path_and_content from 44 lines to 20 lines
All tests pass. Behavior unchanged.
macOS uses U+202F (Narrow No-Break Space) in screenshot filenames
between the time and am/pm. When users type or paste these paths,
they use regular spaces, causing file-not-found errors.
Changes:
- Add resolve_path_with_unicode_fallback() to try U+202F variants
- Add resolve_paths_in_shell_command() for shell command paths
- Apply fix to read_file, read_image, and shell tools
- Fix read_image prompt docs: file_path -> file_paths (array)
- Add 6 unit tests for Unicode space normalization
