5 files changed, 92 insertions(+), 7 deletions(-)

M CLAUDE.md
M README.md
M internal/llmfs/example.go
M internal/llmfs/root.go
M internal/llmfs/stream.go

@@ 313,6 313,15 @@ The following scenarios have been tested and verified working:
 - [x] `echo 'prompt' > /n/llm/ask` followed by `cat /n/llm/ask` - Full LLM interaction works
 - [x] LLM correctly identifies client as Inferno OS when asked
 
+### Streaming (plan9port)
+- [x] `ls stream` - Lists `ask` and `chunk` files
+- [x] `echo "prompt" | 9p write stream/ask` - Starts streaming request
+- [x] `9p read stream/chunk` - Returns streamed chunks
+- [x] Multiple chunks received for longer responses
+- [x] EOF returned when stream completes
+- [x] Short response ("Write a haiku") streams correctly
+- [x] Long response ("Count 1 to 20") streams all content
+
 ## Future Enhancements
 
 - [ ] Multiple conversation support (via subdirectories)

@@ 150,22 150,53 @@ cat /mnt/llm/_example
 ├── context          # Read: conversation history; Write: add system message
 ├── _example         # Read-only: usage examples
 └── stream/          # Streaming interface
-    └── chunk        # Read blocks until next chunk, EOF on completion
+    ├── ask          # Write-only: starts a streaming request
+    └── chunk        # Read-only: blocks until next chunk, EOF on completion
 ```
 
 ### File Behaviors
 
 | File | Read | Write |
 |------|------|-------|
-| `ask` | Returns last LLM response | Sends prompt to LLM, stores response |
+| `ask` | Returns last LLM response | Sends prompt to LLM (sync), stores response |
 | `model` | Returns current model name | Sets model for subsequent requests |
 | `temperature` | Returns current temperature | Sets temperature (0.0-2.0) |
 | `tokens` | Returns last response token count | Permission denied |
 | `new` | Permission denied | Any write resets conversation state |
 | `context` | Returns JSON conversation history | Appends system message to context |
 | `_example` | Returns usage examples | Permission denied |
+| `stream/ask` | Permission denied | Starts a streaming request |
 | `stream/chunk` | Blocks until next chunk, returns it | Permission denied |
 
+## Streaming
+
+For long responses, use the streaming interface to see output as it's generated:
+
+```bash
+# Start a streaming request (using 9p tool)
+echo "Write a poem about the moon" | 9p -a localhost:5640 write stream/ask &
+
+# Read chunks as they arrive
+while chunk=$(9p -a localhost:5640 read stream/chunk 2>/dev/null); do
+  [ -z "$chunk" ] && break
+  printf "%s" "$chunk"
+done
+```
+
+With a mounted filesystem:
+
+```bash
+# Start streaming in background
+echo "Explain quantum computing" > /mnt/llm/stream/ask &
+
+# Read chunks
+while read -r chunk < /mnt/llm/stream/chunk 2>/dev/null; do
+  printf "%s" "$chunk"
+done
+```
+
+**Note:** Start reading chunks immediately after writing to `stream/ask`. If you wait too long, the stream may complete and you'll get EOF.
+
 ## Shell Scripting
 
 ```bash

@@ 25,9 25,11 @@ Conversation Management:
 Token Usage:
   cat tokens                     # View tokens from last response
 
-Streaming (Advanced):
-  echo "Tell me a story" > ask   # Start generating
-  cat stream/chunk               # Read chunks as they arrive (blocks)
+Streaming:
+  echo "Tell me a story" > stream/ask  # Start streaming request
+  cat stream/chunk                      # Read next chunk (blocks until available)
+  # Keep reading stream/chunk until EOF for full response
+  # Note: Read chunks immediately after writing to stream/ask
 
 Shell Scripting:
   #!/bin/sh


@@ 46,14 48,15 @@ Environment:
   ANTHROPIC_API_KEY must be set when starting the server
 
 Files:
-  ask          Read/write: prompt goes in, response comes out
+  ask          Read/write: prompt goes in, response comes out (sync)
   model        Read/write: current model name
   temperature  Read/write: sampling temperature (0.0-2.0)
   tokens       Read-only: token count from last response
   new          Write-only: any write resets conversation
   context      Read: JSON history; Write: add system message
   _example     Read-only: this help text
-  stream/chunk Read-only: streaming chunks (blocking)
+  stream/ask   Write-only: starts a streaming request
+  stream/chunk Read-only: returns next chunk (blocks), EOF when done
 `
 
 // NewExampleFile creates the _example file with usage examples

@@ 21,6 21,7 @@ func NewRoot(client *llm.Client) protocol.Dir {
 
 	// Add stream directory
 	streamDir := protocol.NewStaticDir("stream")
+	streamDir.AddChild(NewStreamAskFile(client))
 	streamDir.AddChild(NewChunkFile(client))
 	root.AddChild(streamDir)
 

@@ 1,7 1,9 @@
 package llmfs
 
 import (
+	"context"
 	"io"
+	"strings"
 
 	"github.com/NERVsystems/llm9p/internal/llm"
 	"github.com/NERVsystems/llm9p/internal/protocol"


@@ 51,3 53,42 @@ func (f *ChunkFile) Stat() protocol.Stat {
 	s.Length = 0
 	return s
 }
+
+// StreamAskFile starts a streaming request
+// Write a prompt to start streaming, then read chunks from stream/chunk
+type StreamAskFile struct {
+	*protocol.BaseFile
+	client *llm.Client
+}
+
+// NewStreamAskFile creates the stream/ask file
+func NewStreamAskFile(client *llm.Client) *StreamAskFile {
+	return &StreamAskFile{
+		BaseFile: protocol.NewBaseFile("ask", 0222), // write-only
+		client:   client,
+	}
+}
+
+func (f *StreamAskFile) Read(p []byte, offset int64) (int, error) {
+	return 0, protocol.ErrPermission
+}
+
+func (f *StreamAskFile) Write(p []byte, offset int64) (int, error) {
+	prompt := strings.TrimSpace(string(p))
+	if prompt == "" {
+		return len(p), nil
+	}
+
+	// Start streaming - chunks will be available via stream/chunk
+	err := f.client.StartStream(context.Background(), prompt)
+	if err != nil {
+		// Return error to indicate stream failed to start
+		return 0, err
+	}
+
+	return len(p), nil
+}
+
+func (f *StreamAskFile) Stat() protocol.Stat {
+	return f.BaseFile.Stat()
+}